fix .csv to .csv.gz transition

This reverts commit ccf55b2308.

.github/workflows/adsb-to-aircraft-for-day.yaml

@@ -1,182 +0,0 @@
-name: Historical ADS-B Processing
-
-on:
-  workflow_dispatch:
-    inputs:
-      date:
-        description: 'YYYY-MM-DD'
-        required: true
-        type: string
-      concat_with_latest_csv:
-        description: 'Also concatenate with latest CSV from GitHub releases'
-        required: false
-        type: boolean
-        default: false
-  workflow_call:
-    inputs:
-      date:
-        description: 'YYYY-MM-DD'
-        required: true
-        type: string
-      concat_with_latest_csv:
-        description: 'Also concatenate with latest CSV from GitHub releases'
-        required: false
-        type: boolean
-        default: false
-
-jobs:
-  adsb-extract:
-    runs-on: ubuntu-24.04-arm
-    steps:
-      - name: Checkout
-        uses: actions/checkout@v6
-
-      - name: Setup Python
-        uses: actions/setup-python@v6
-        with:
-          python-version: '3.12'
-
-      - name: Install dependencies
-        run: |
-          python -m pip install --upgrade pip
-          pip install -r requirements.txt
-
-      - name: Download and split ADS-B data
-        env:
-          DATE: ${{ inputs.date }}
-          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-        run: |
-          python -m src.adsb.download_and_list_icaos --date "$DATE"
-          ls -lah data/output/adsb_archives/"$DATE" || true
-
-      - name: Upload archive part 0
-        uses: actions/upload-artifact@v4
-        with:
-          name: adsb-archive-${{ inputs.date }}-part-0
-          path: data/output/adsb_archives/${{ inputs.date }}/${{ inputs.date }}_part_0.tar.gz
-          retention-days: 1
-          compression-level: 0
-          if-no-files-found: error
-
-      - name: Upload archive part 1
-        uses: actions/upload-artifact@v4
-        with:
-          name: adsb-archive-${{ inputs.date }}-part-1
-          path: data/output/adsb_archives/${{ inputs.date }}/${{ inputs.date }}_part_1.tar.gz
-          retention-days: 1
-          compression-level: 0
-          if-no-files-found: error
-
-      - name: Upload archive part 2
-        uses: actions/upload-artifact@v4
-        with:
-          name: adsb-archive-${{ inputs.date }}-part-2
-          path: data/output/adsb_archives/${{ inputs.date }}/${{ inputs.date }}_part_2.tar.gz
-          retention-days: 1
-          compression-level: 0
-          if-no-files-found: error
-
-      - name: Upload archive part 3
-        uses: actions/upload-artifact@v4
-        with:
-          name: adsb-archive-${{ inputs.date }}-part-3
-          path: data/output/adsb_archives/${{ inputs.date }}/${{ inputs.date }}_part_3.tar.gz
-          retention-days: 1
-          compression-level: 0
-          if-no-files-found: error
-
-  adsb-map:
-    needs: adsb-extract
-    runs-on: ubuntu-24.04-arm
-    strategy:
-      fail-fast: true
-      matrix:
-        part_id: [0, 1, 2, 3]
-    steps:
-      - name: Checkout
-        uses: actions/checkout@v6
-
-      - name: Setup Python
-        uses: actions/setup-python@v6
-        with:
-          python-version: '3.12'
-
-      - name: Install dependencies
-        run: |
-          python -m pip install --upgrade pip
-          pip install -r requirements.txt
-
-      - name: Download archive part
-        uses: actions/download-artifact@v4
-        with:
-          name: adsb-archive-${{ inputs.date }}-part-${{ matrix.part_id }}
-          path: data/output/adsb_archives/${{ inputs.date }}
-
-      - name: Verify archive
-        run: |
-          FILE="data/output/adsb_archives/${{ inputs.date }}/${{ inputs.date }}_part_${{ matrix.part_id }}.tar.gz"
-          ls -lah data/output/adsb_archives/${{ inputs.date }}/
-          if [ ! -f "$FILE" ]; then
-            echo "::error::Archive not found: $FILE"
-            exit 1
-          fi
-          echo "Verified: $(du -h "$FILE")"
-
-      - name: Process part
-        env:
-          DATE: ${{ inputs.date }}
-        run: |
-          python -m src.adsb.process_icao_chunk --part-id ${{ matrix.part_id }} --date "$DATE"
-
-      - name: Upload compressed outputs
-        uses: actions/upload-artifact@v4
-        with:
-          name: adsb-compressed-${{ inputs.date }}-part-${{ matrix.part_id }}
-          path: data/output/compressed/${{ inputs.date }}
-          retention-days: 1
-          compression-level: 0
-          if-no-files-found: error
-
-  adsb-reduce:
-    needs: adsb-map
-    runs-on: ubuntu-24.04-arm
-    steps:
-      - name: Checkout
-        uses: actions/checkout@v6
-
-      - name: Setup Python
-        uses: actions/setup-python@v6
-        with:
-          python-version: '3.12'
-
-      - name: Install dependencies
-        run: |
-          python -m pip install --upgrade pip
-          pip install -r requirements.txt
-
-      - name: Download compressed outputs
-        uses: actions/download-artifact@v4
-        with:
-          pattern: adsb-compressed-${{ inputs.date }}-part-*
-          path: data/output/compressed/${{ inputs.date }}
-          merge-multiple: true
-
-      - name: Concatenate final outputs
-        env:
-          DATE: ${{ inputs.date }}
-          CONCAT_WITH_LATEST_CSV: ${{ inputs.concat_with_latest_csv }}
-        run: |
-          EXTRA=""
-          if [ "$CONCAT_WITH_LATEST_CSV" = "true" ]; then
-            EXTRA="--concat_with_latest_csv"
-          fi
-          python -m src.adsb.concat_parquet_to_final --date "$DATE" $EXTRA
-          ls -lah data/output/ || true
-
-      - name: Upload final artifacts
-        uses: actions/upload-artifact@v4
-        with:
-          name: openairframes_adsb-${{ inputs.date }}
-          path: data/output/openairframes_adsb_*
-          retention-days: 30
-          if-no-files-found: error

.github/workflows/adsb-to-aircraft-multiple-day-run.yaml

@@ -1,118 +0,0 @@
-name: adsb-to-aircraft-multiple-day-run
-
-on:
-  workflow_dispatch:
-    inputs:
-      start_date:
-        description: 'YYYY-MM-DD (inclusive)'
-        required: true
-        type: string
-      end_date:
-        description: 'YYYY-MM-DD (exclusive)'
-        required: true
-        type: string
-
-jobs:
-  generate-dates:
-    runs-on: ubuntu-24.04-arm
-    outputs:
-      dates: ${{ steps.generate.outputs.dates }}
-    steps:
-      - name: Generate date list
-        id: generate
-        env:
-          START_DATE: ${{ inputs.start_date }}
-          END_DATE: ${{ inputs.end_date }}
-        run: |
-          python - <<'PY'
-          import json
-          import os
-          from datetime import datetime, timedelta
-
-          start = datetime.strptime(os.environ["START_DATE"], "%Y-%m-%d")
-          end = datetime.strptime(os.environ["END_DATE"], "%Y-%m-%d")
-          if end <= start:
-            raise SystemExit("end_date must be after start_date")
-
-          dates = []
-          cur = start
-          while cur < end:
-            dates.append(cur.strftime("%Y-%m-%d"))
-            cur += timedelta(days=1)
-
-          with open(os.environ["GITHUB_OUTPUT"], "a") as f:
-            f.write(f"dates={json.dumps(dates)}\n")
-          PY
-
-  adsb-day:
-    needs: generate-dates
-    strategy:
-      fail-fast: true
-      matrix:
-        date: ${{ fromJson(needs.generate-dates.outputs.dates) }}
-    uses: ./.github/workflows/adsb-to-aircraft-for-day.yaml
-    with:
-      date: ${{ matrix.date }}
-
-  adsb-final:
-    needs: adsb-day
-    runs-on: ubuntu-24.04-arm
-    steps:
-      - name: Checkout
-        uses: actions/checkout@v6
-
-      - name: Setup Python
-        uses: actions/setup-python@v6
-        with:
-          python-version: '3.12'
-
-      - name: Install dependencies
-        run: |
-          python -m pip install --upgrade pip
-          pip install -r requirements.txt
-
-      - name: Download daily CSVs
-        uses: actions/download-artifact@v4
-        with:
-          pattern: openairframes_adsb-*
-          path: outputs/daily/
-          merge-multiple: true
-
-      - name: Concatenate all days to final CSV
-        env:
-          START_DATE: ${{ inputs.start_date }}
-          END_DATE: ${{ inputs.end_date }}
-        run: |
-          python - <<'PY'
-          import os
-          import re
-          from pathlib import Path
-          import polars as pl
-
-          start = os.environ["START_DATE"]
-          end = os.environ["END_DATE"]
-          daily_dir = Path("outputs/daily")
-          files = sorted(daily_dir.glob("openairframes_adsb_*.csv.gz"))
-          if not files:
-            raise SystemExit("No daily CSVs found")
-
-          def date_key(path: Path) -> str:
-            m = re.match(r"openairframes_adsb_(\d{4}-\d{2}-\d{2})_", path.name)
-            return m.group(1) if m else path.name
-
-          files = sorted(files, key=date_key)
-          frames = [pl.read_csv(p) for p in files]
-          df = pl.concat(frames, how="vertical", rechunk=True)
-
-          output_path = Path("outputs") / f"openairframes_adsb_{start}_{end}.csv.gz"
-          df.write_csv(output_path, compression="gzip")
-          print(f"Wrote {output_path} with {df.height} rows")
-          PY
-
-      - name: Upload final CSV
-        uses: actions/upload-artifact@v4
-        with:
-          name: openairframes_adsb-${{ inputs.start_date }}-${{ inputs.end_date }}
-          path: outputs/openairframes_adsb_${{ inputs.start_date }}_${{ inputs.end_date }}.csv.gz
-          retention-days: 30
-# gh workflow run adsb-to-aircraft-multiple-day-run.yaml --repo ggman12/OpenAirframes --ref jonah/fix-historical-proper -f start_date=2025-12-31 -f end_date=2026-01-02

.github/workflows/historical-adsb.yaml

@@ -0,0 +1,286 @@
+name: Historical ADS-B Processing
+
+on:
+  workflow_dispatch:
+    inputs:
+      start_date:
+        description: 'Start date (YYYY-MM-DD, inclusive)'
+        required: true
+        type: string
+      end_date:
+        description: 'End date (YYYY-MM-DD, exclusive)'
+        required: true
+        type: string
+      chunk_days:
+        description: 'Days per job chunk (default: 7)'
+        required: false
+        type: number
+        default: 7
+
+jobs:
+  generate-matrix:
+    runs-on: ubuntu-latest
+    outputs:
+      chunks: ${{ steps.generate.outputs.chunks }}
+      global_start: ${{ inputs.start_date }}
+      global_end: ${{ inputs.end_date }}
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Setup Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.12'
+
+      - name: Generate date chunks
+        id: generate
+        env:
+          INPUT_START_DATE: ${{ inputs.start_date }}
+          INPUT_END_DATE: ${{ inputs.end_date }}
+          INPUT_CHUNK_DAYS: ${{ inputs.chunk_days }}
+        run: python src/adsb/historical_generate_matrix.py
+
+  adsb-extract:
+    needs: generate-matrix
+    runs-on: ubuntu-24.04-arm
+    strategy:
+      matrix:
+        chunk: ${{ fromJson(needs.generate-matrix.outputs.chunks) }}
+      max-parallel: 3
+      fail-fast: true
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Setup Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.12'
+
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          pip install -r requirements.txt
+
+      - name: Free disk space
+        run: |
+          sudo rm -rf /usr/share/dotnet
+          sudo rm -rf /opt/ghc
+          sudo rm -rf /usr/local/share/boost
+          df -h
+
+      - name: Download and extract ADS-B data
+        env:
+          START_DATE: ${{ matrix.chunk.start_date }}
+          END_DATE: ${{ matrix.chunk.end_date }}
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: |
+          python -m src.adsb.download_and_list_icaos --start-date "$START_DATE" --end-date "$END_DATE"
+          ls -lah data/output/
+
+      - name: Create tar of extracted data and split into chunks
+        run: |
+          cd data/output
+          echo "=== Disk space before tar ==="
+          df -h .
+          echo "=== Files to tar ==="
+          ls -lah *-planes-readsb-prod-0.tar_0 icao_manifest_*.txt 2>/dev/null || echo "No files found"
+          
+          # Create tar with explicit error checking
+          if ls *-planes-readsb-prod-0.tar_0 1>/dev/null 2>&1; then
+            tar -cvf extracted_data.tar *-planes-readsb-prod-0.tar_0 icao_manifest_*.txt
+            echo "=== Tar file created ==="
+            ls -lah extracted_data.tar
+            # Verify tar integrity
+            tar -tf extracted_data.tar > /dev/null && echo "Tar integrity check passed" || { echo "Tar integrity check FAILED"; exit 1; }
+            
+            # Record tar size and checksum for verification after reassembly
+            echo "=== Recording tar metadata ==="
+            ORIGINAL_SIZE=$(stat --format=%s extracted_data.tar)
+            ORIGINAL_SHA=$(sha256sum extracted_data.tar | awk '{print $1}')
+            echo "Size: $ORIGINAL_SIZE"
+            echo "SHA256: $ORIGINAL_SHA"
+            
+            # Split into 500MB chunks to avoid artifact upload issues
+            echo "=== Splitting tar into 500MB chunks ==="
+            mkdir -p tar_chunks
+            split -b 500M extracted_data.tar tar_chunks/extracted_data.tar.part_
+            rm extracted_data.tar
+            
+            # Write metadata file (plain text so artifact upload won't skip it)
+            echo "$ORIGINAL_SHA  extracted_data.tar" > tar_chunks/checksum.txt
+            echo "$ORIGINAL_SIZE" >> tar_chunks/checksum.txt
+            
+            echo "=== Chunks created ==="
+            ls -lah tar_chunks/
+            echo "=== Checksum file ==="
+            cat tar_chunks/checksum.txt
+          else
+            echo "ERROR: No extracted directories found, cannot create tar"
+            exit 1
+          fi
+
+      - name: Upload extracted data chunks
+        uses: actions/upload-artifact@v4
+        with:
+          name: adsb-extracted-${{ matrix.chunk.start_date }}-${{ matrix.chunk.end_date }}
+          path: data/output/tar_chunks/
+          retention-days: 1
+          compression-level: 0
+          if-no-files-found: warn
+
+  adsb-map:
+    needs: [generate-matrix, adsb-extract]
+    runs-on: ubuntu-24.04-arm
+    strategy:
+      fail-fast: true
+      matrix:
+        chunk: ${{ fromJson(needs.generate-matrix.outputs.chunks) }}
+        icao_chunk: [0, 1, 2, 3]
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Setup Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.12'
+
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          pip install -r requirements.txt
+
+      - name: Free disk space
+        run: |
+          sudo rm -rf /usr/share/dotnet
+          sudo rm -rf /opt/ghc
+          sudo rm -rf /usr/local/share/boost
+          df -h
+
+      - name: Download extracted data
+        uses: actions/download-artifact@v4
+        with:
+          name: adsb-extracted-${{ matrix.chunk.start_date }}-${{ matrix.chunk.end_date }}
+          path: data/output/tar_chunks/
+
+      - name: Reassemble and extract tar
+        id: extract
+        run: |
+          cd data/output
+          if [ -d tar_chunks ] && ls tar_chunks/extracted_data.tar.part_* 1>/dev/null 2>&1; then
+            echo "=== Chunk files info ==="
+            ls -lah tar_chunks/
+            
+            cd tar_chunks
+            
+            # Reassemble tar with explicit sorting
+            echo "=== Reassembling tar file ==="
+            ls -1 extracted_data.tar.part_?? | sort | while read part; do
+              echo "Appending $part..."
+              cat "$part" >> ../extracted_data.tar
+            done
+            cd ..
+            
+            echo "=== Reassembled tar file info ==="
+            ls -lah extracted_data.tar
+            
+            # Verify integrity
+            echo "=== Verifying reassembled tar ==="
+            if [ -f tar_chunks/checksum.txt ]; then
+              EXPECTED_SHA=$(head -1 tar_chunks/checksum.txt | awk '{print $1}')
+              EXPECTED_SIZE=$(sed -n '2p' tar_chunks/checksum.txt)
+              ACTUAL_SHA=$(sha256sum extracted_data.tar | awk '{print $1}')
+              ACTUAL_SIZE=$(stat --format=%s extracted_data.tar)
+              echo "Expected: SHA=$EXPECTED_SHA Size=$EXPECTED_SIZE"
+              echo "Actual:   SHA=$ACTUAL_SHA Size=$ACTUAL_SIZE"
+              if [ "$EXPECTED_SHA" != "$ACTUAL_SHA" ] || [ "$EXPECTED_SIZE" != "$ACTUAL_SIZE" ]; then
+                echo "ERROR: Reassembled tar does not match original - data corrupted during transfer"
+                exit 1
+              fi
+              echo "Checksum and size verified"
+            else
+              echo "WARNING: No checksum file found, falling back to tar integrity check"
+              tar -tf extracted_data.tar > /dev/null || { echo "ERROR: Tar file is corrupted"; exit 1; }
+              echo "Tar integrity check passed"
+            fi
+            
+            rm -rf tar_chunks
+            
+            echo "=== Extracting ==="
+            tar -xf extracted_data.tar
+            rm extracted_data.tar
+            echo "has_data=true" >> "$GITHUB_OUTPUT"
+            echo "=== Contents of data/output ==="
+            ls -lah
+          else
+            echo "No tar chunks found"
+            echo "has_data=false" >> "$GITHUB_OUTPUT"
+          fi
+
+      - name: Process ICAO chunk
+        if: steps.extract.outputs.has_data == 'true'
+        env:
+          START_DATE: ${{ matrix.chunk.start_date }}
+          END_DATE: ${{ matrix.chunk.end_date }}
+        run: |
+          python -m src.adsb.process_icao_chunk --chunk-id ${{ matrix.icao_chunk }} --total-chunks 4 --start-date "$START_DATE" --end-date "$END_DATE"
+          ls -lah data/output/adsb_chunks/ || echo "No chunks created"
+
+      - name: Upload chunk artifacts
+        if: steps.extract.outputs.has_data == 'true'
+        uses: actions/upload-artifact@v4
+        with:
+          name: adsb-map-${{ matrix.chunk.start_date }}-${{ matrix.chunk.end_date }}-chunk-${{ matrix.icao_chunk }}
+          path: data/output/adsb_chunks/
+          retention-days: 1
+          if-no-files-found: ignore
+
+  adsb-reduce:
+    needs: [generate-matrix, adsb-map]
+    runs-on: ubuntu-24.04-arm
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Setup Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.12'
+
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          pip install -r requirements.txt
+
+      - name: Download all chunk artifacts
+        uses: actions/download-artifact@v4
+        with:
+          pattern: adsb-map-*
+          path: data/output/adsb_chunks/
+          merge-multiple: true
+
+      - name: Debug downloaded files
+        run: |
+          echo "=== Disk space before processing ==="
+          df -h
+          echo "=== Listing data/output/adsb_chunks/ ==="
+          find data/output/adsb_chunks/ -type f 2>/dev/null | wc -l
+          echo "=== Total parquet size ==="
+          du -sh data/output/adsb_chunks/ || echo "No chunks dir"
+
+      - name: Combine chunks to CSV
+        env:
+          START_DATE: ${{ needs.generate-matrix.outputs.global_start }}
+          END_DATE: ${{ needs.generate-matrix.outputs.global_end }}
+        run: |
+          python -m src.adsb.combine_chunks_to_csv --chunks-dir data/output/adsb_chunks --start-date "$START_DATE" --end-date "$END_DATE" --skip-base --stream
+          ls -lah data/openairframes/
+
+      - name: Upload final artifact
+        uses: actions/upload-artifact@v4
+        with:
+          name: openairframes_adsb-${{ needs.generate-matrix.outputs.global_start }}-${{ needs.generate-matrix.outputs.global_end }}
+          path: data/openairframes/*.csv.gz
+          retention-days: 30

.github/workflows/openairframes-daily-release.yaml

@@ -1,4 +1,4 @@
-name: openairframes-daily-release
+name: OpenAirframes Daily Release
 
 on:
   schedule:
@@ -76,75 +76,159 @@ jobs:
             data/faa_releasable/ReleasableAircraft_*.zip
           retention-days: 1
 
-  resolve-dates:
-    runs-on: ubuntu-latest
+  adsb-extract:
+    runs-on: ubuntu-24.04-arm
     if: github.event_name != 'schedule'
     outputs:
-      date: ${{ steps.out.outputs.date }}
-      adsb_date: ${{ steps.out.outputs.adsb_date }}
-    steps:
-      - id: out
-        run: |
-          if [ -n "${{ inputs.date }}" ]; then
-            echo "date=${{ inputs.date }}" >> "$GITHUB_OUTPUT"
-            echo "adsb_date=${{ inputs.date }}" >> "$GITHUB_OUTPUT"
-          else
-            echo "date=$(date -u -d 'yesterday' +%Y-%m-%d)" >> "$GITHUB_OUTPUT"
-            echo "adsb_date=$(date -u -d 'yesterday' +%Y-%m-%d)" >> "$GITHUB_OUTPUT"
-          fi
-
-  adsb-to-aircraft:
-    needs: resolve-dates
-    if: github.event_name != 'schedule'
-    uses: ./.github/workflows/adsb-to-aircraft-for-day.yaml
-    with:
-      date: ${{ needs.resolve-dates.outputs.adsb_date }}
-      concat_with_latest_csv: true
-
-  adsb-reduce:
-    needs: [resolve-dates, adsb-to-aircraft]
-    if: always() && github.event_name != 'schedule' && needs.adsb-to-aircraft.result == 'failure'
-    runs-on: ubuntu-24.04-arm
+      manifest-exists: ${{ steps.check.outputs.exists }}
     steps:
       - name: Checkout
         uses: actions/checkout@v6
+        with:
+          fetch-depth: 0
 
       - name: Setup Python
         uses: actions/setup-python@v6
         with:
-          python-version: '3.12'
+          python-version: "3.14"
 
       - name: Install dependencies
         run: |
           python -m pip install --upgrade pip
           pip install -r requirements.txt
 
-      - name: Download compressed outputs
-        uses: actions/download-artifact@v4
-        with:
-          pattern: adsb-compressed-${{ needs.resolve-dates.outputs.adsb_date }}-part-*
-          path: data/output/compressed/${{ needs.resolve-dates.outputs.adsb_date }}
-          merge-multiple: true
-
-      - name: Concatenate final outputs
+      - name: Download and extract ADS-B data
         env:
-          DATE: ${{ needs.resolve-dates.outputs.adsb_date }}
-          CONCAT_WITH_LATEST_CSV: true
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
         run: |
-          EXTRA=""
-          if [ "$CONCAT_WITH_LATEST_CSV" = "true" ]; then
-            EXTRA="--concat_with_latest_csv"
-          fi
-          python -m src.adsb.concat_parquet_to_final --date "$DATE" $EXTRA
-          ls -lah data/output/ || true
+          python -m src.adsb.download_and_list_icaos ${{ inputs.date && format('--date {0}', inputs.date) || '' }}
+          ls -lah data/output/
 
-      - name: Upload final artifacts
+      - name: Check manifest exists
+        id: check
+        run: |
+          if ls data/output/icao_manifest_*.txt 1>/dev/null 2>&1; then
+            echo "exists=true" >> "$GITHUB_OUTPUT"
+          else
+            echo "exists=false" >> "$GITHUB_OUTPUT"
+          fi
+
+      - name: Create tar of extracted data
+        run: |
+          cd data/output
+          tar -cf extracted_data.tar *-planes-readsb-prod-0.tar_0 icao_manifest_*.txt
+          ls -lah extracted_data.tar
+
+      - name: Upload extracted data
         uses: actions/upload-artifact@v4
         with:
-          name: openairframes_adsb-${{ needs.resolve-dates.outputs.adsb_date }}
-          path: data/output/openairframes_adsb_*
-          retention-days: 30
-          if-no-files-found: error
+          name: adsb-extracted
+          path: data/output/extracted_data.tar
+          retention-days: 1
+          compression-level: 0  # Already compressed trace files
+
+  adsb-map:
+    runs-on: ubuntu-24.04-arm
+    needs: adsb-extract
+    if: github.event_name != 'schedule' && needs.adsb-extract.outputs.manifest-exists == 'true'
+    strategy:
+      fail-fast: false
+      matrix:
+        chunk: [0, 1, 2, 3]
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v6
+        with:
+          fetch-depth: 0
+
+      - name: Setup Python
+        uses: actions/setup-python@v6
+        with:
+          python-version: "3.14"
+
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          pip install -r requirements.txt
+
+      - name: Download extracted data
+        uses: actions/download-artifact@v4
+        with:
+          name: adsb-extracted
+          path: data/output/
+
+      - name: Extract tar
+        run: |
+          cd data/output
+          tar -xf extracted_data.tar
+          rm extracted_data.tar
+          echo "=== Contents of data/output ==="
+          ls -lah
+          echo "=== Looking for manifest ==="
+          cat icao_manifest_*.txt | head -20 || echo "No manifest found"
+          echo "=== Looking for extracted dirs ==="
+          ls -d *-planes-readsb-prod-0* 2>/dev/null || echo "No extracted dirs"
+
+      - name: Process chunk ${{ matrix.chunk }}
+        run: |
+          python -m src.adsb.process_icao_chunk --chunk-id ${{ matrix.chunk }} --total-chunks 4 ${{ inputs.date && format('--date {0}', inputs.date) || '' }}
+          mkdir -p data/output/adsb_chunks
+          ls -lah data/output/adsb_chunks/ || echo "No chunks created"
+
+      - name: Upload chunk artifacts
+        uses: actions/upload-artifact@v4
+        with:
+          name: adsb-chunk-${{ matrix.chunk }}
+          path: data/output/adsb_chunks/
+          retention-days: 1
+
+  adsb-reduce:
+    runs-on: ubuntu-24.04-arm
+    needs: adsb-map
+    if: github.event_name != 'schedule'
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v6
+        with:
+          fetch-depth: 0
+
+      - name: Setup Python
+        uses: actions/setup-python@v6
+        with:
+          python-version: "3.14"
+
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          pip install -r requirements.txt
+
+      - name: Download all chunk artifacts
+        uses: actions/download-artifact@v4
+        with:
+          pattern: adsb-chunk-*
+          path: data/output/adsb_chunks/
+          merge-multiple: true
+
+      - name: Debug downloaded files
+        run: |
+          echo "=== Listing data/ ==="
+          find data/ -type f 2>/dev/null | head -50 || echo "No files in data/"
+          echo "=== Looking for parquet files ==="
+          find . -name "*.parquet" 2>/dev/null | head -20 || echo "No parquet files found"
+
+      - name: Combine chunks to CSV
+        run: |
+          mkdir -p data/output/adsb_chunks
+          ls -lah data/output/adsb_chunks/ || echo "Directory empty or does not exist"
+          python -m src.adsb.combine_chunks_to_csv --chunks-dir data/output/adsb_chunks ${{ inputs.date && format('--date {0}', inputs.date) || '' }}
+          ls -lah data/openairframes/
+
+      - name: Upload ADS-B artifacts
+        uses: actions/upload-artifact@v4
+        with:
+          name: adsb-release
+          path: data/openairframes/openairframes_adsb_*.csv.gz
+          retention-days: 1
 
   build-community:
     runs-on: ubuntu-latest
@@ -233,14 +317,9 @@ jobs:
 
   create-release:
     runs-on: ubuntu-latest
-    needs: [resolve-dates, build-faa, adsb-to-aircraft, adsb-reduce, build-community, build-adsbexchange-json, build-mictronics-db]
-    if: github.event_name != 'schedule' && !cancelled()
+    needs: [build-faa, adsb-reduce, build-community, build-adsbexchange-json, build-mictronics-db]
+    if: github.event_name != 'schedule' && !failure() && !cancelled()
     steps:
-      - name: Check ADS-B workflow status
-        if: needs.adsb-to-aircraft.result != 'success' && needs.adsb-reduce.result != 'success'
-        run: |
-          echo "WARNING: ADS-B workflow failed (adsb-to-aircraft='${{ needs.adsb-to-aircraft.result }}', adsb-reduce='${{ needs.adsb-reduce.result }}'), will continue without ADS-B artifacts"
-
       - name: Checkout for gh CLI
         uses: actions/checkout@v4
         with:
@@ -249,33 +328,31 @@ jobs:
           sparse-checkout-cone-mode: false
 
       - name: Download FAA artifacts
-        uses: actions/download-artifact@v5
+        uses: actions/download-artifact@v4
         with:
           name: faa-release
           path: artifacts/faa
 
       - name: Download ADS-B artifacts
-        uses: actions/download-artifact@v5
-        if: needs.adsb-to-aircraft.result == 'success' || needs.adsb-reduce.result == 'success'
-        continue-on-error: true
+        uses: actions/download-artifact@v4
         with:
-          name: openairframes_adsb-${{ needs.resolve-dates.outputs.adsb_date }}
+          name: adsb-release
           path: artifacts/adsb
 
       - name: Download Community artifacts
-        uses: actions/download-artifact@v5
+        uses: actions/download-artifact@v4
         with:
           name: community-release
           path: artifacts/community
 
       - name: Download ADS-B Exchange JSON artifact
-        uses: actions/download-artifact@v5
+        uses: actions/download-artifact@v4
         with:
           name: adsbexchange-json
           path: artifacts/adsbexchange
 
       - name: Download Mictronics DB artifact
-        uses: actions/download-artifact@v5
+        uses: actions/download-artifact@v4
         continue-on-error: true
         with:
           name: mictronics-db
@@ -311,11 +388,7 @@ jobs:
           
           # Find files from artifacts using find (handles nested structures)
           CSV_FILE_FAA=$(find artifacts/faa -name "openairframes_faa_*.csv" -type f 2>/dev/null | head -1)
-          # Prefer concatenated file (with date range) over single-day file
-          CSV_FILE_ADSB=$(find artifacts/adsb -name "openairframes_adsb_*_*.csv.gz" -type f 2>/dev/null | head -1)
-          if [ -z "$CSV_FILE_ADSB" ]; then
-            CSV_FILE_ADSB=$(find artifacts/adsb -name "openairframes_adsb_*.csv.gz" -type f 2>/dev/null | head -1)
-          fi
+          CSV_FILE_ADSB=$(find artifacts/adsb -name "openairframes_adsb_*.csv.gz" -type f 2>/dev/null | head -1)
           CSV_FILE_COMMUNITY=$(find artifacts/community -name "openairframes_community_*.csv" -type f 2>/dev/null | head -1)
           ZIP_FILE=$(find artifacts/faa -name "ReleasableAircraft_*.zip" -type f 2>/dev/null | head -1)
           JSON_FILE_ADSBX=$(find artifacts/adsbexchange -name "basic-ac-db_*.json.gz" -type f 2>/dev/null | head -1)
@@ -326,6 +399,9 @@ jobs:
           if [ -z "$CSV_FILE_FAA" ] || [ ! -f "$CSV_FILE_FAA" ]; then
             MISSING_FILES="$MISSING_FILES FAA_CSV"
           fi
+          if [ -z "$CSV_FILE_ADSB" ] || [ ! -f "$CSV_FILE_ADSB" ]; then
+            MISSING_FILES="$MISSING_FILES ADSB_CSV"
+          fi
           if [ -z "$ZIP_FILE" ] || [ ! -f "$ZIP_FILE" ]; then
             MISSING_FILES="$MISSING_FILES FAA_ZIP"
           fi
@@ -335,11 +411,6 @@ jobs:
           
           # Optional files - warn but don't fail
           OPTIONAL_MISSING=""
-          if [ -z "$CSV_FILE_ADSB" ] || [ ! -f "$CSV_FILE_ADSB" ]; then
-            OPTIONAL_MISSING="$OPTIONAL_MISSING ADSB_CSV"
-            CSV_FILE_ADSB=""
-            CSV_BASENAME_ADSB=""
-          fi
           if [ -z "$ZIP_FILE_MICTRONICS" ] || [ ! -f "$ZIP_FILE_MICTRONICS" ]; then
             OPTIONAL_MISSING="$OPTIONAL_MISSING MICTRONICS_ZIP"
             ZIP_FILE_MICTRONICS=""
@@ -357,9 +428,7 @@ jobs:
           
           # Get basenames for display
           CSV_BASENAME_FAA=$(basename "$CSV_FILE_FAA")
-          if [ -n "$CSV_FILE_ADSB" ]; then
-            CSV_BASENAME_ADSB=$(basename "$CSV_FILE_ADSB")
-          fi
+          CSV_BASENAME_ADSB=$(basename "$CSV_FILE_ADSB")
           CSV_BASENAME_COMMUNITY=$(basename "$CSV_FILE_COMMUNITY" 2>/dev/null || echo "")
           ZIP_BASENAME=$(basename "$ZIP_FILE")
           JSON_BASENAME_ADSBX=$(basename "$JSON_FILE_ADSBX")
@@ -414,7 +483,7 @@ jobs:
 
             Assets:
             - ${{ steps.meta.outputs.csv_basename_faa }}
-            ${{ steps.meta.outputs.csv_basename_adsb && format('- {0}', steps.meta.outputs.csv_basename_adsb) || '' }}
+            - ${{ steps.meta.outputs.csv_basename_adsb }}
             - ${{ steps.meta.outputs.csv_basename_community }}
             - ${{ steps.meta.outputs.zip_basename }}
             - ${{ steps.meta.outputs.json_basename_adsbx }}

.gitignore

@@ -281,7 +281,4 @@ read*lock
 .nx/
 
 # jsii-rosetta files
-type-fingerprints.txt
-
-notebooks/whatever.ipynb
-.snapshots/
+type-fingerprints.txt

README.md

@@ -16,19 +16,11 @@ A daily release is created at **06:00 UTC** and includes:
 - **openairframes_community.csv**  
   All community submissions
 
-- **openairframes_adsb.csv**  
-  Airframes dataset derived from ADSB.lol network data. For each UTC day, a row is created for every icao observed in that day’s ADS-B messages, using registration data from [tar1090-db](https://github.com/wiedehopf/tar1090-db) (ADSBExchange & Mictronics).
-Example Usage:
-```python
-import pandas as pd
-url = "https://github.com/PlaneQuery/OpenAirframes/releases/download/openairframes-2026-03-18-main/openairframes_adsb_2024-01-01_2026-03-17.csv.gz" # 1GB
-df = pd.read_csv(url)
-df
-```
-![](docs/images/df_adsb_example_0.png)
 - **openairframes_faa.csv**  
   All [FAA registration data](https://www.faa.gov/licenses_certificates/aircraft_certification/aircraft_registry/releasable_aircraft_download) from 2023-08-16 to present (~260 MB)
 
+- **openairframes_adsb.csv**  
+  Airframe information derived from ADS-B messages on the [ADSB.lol](https://www.adsb.lol/) network, from 2026-02-12 to present (will be from 2024-01-01 soon). The airframe information originates from [mictronics aircraft database](https://www.mictronics.de/aircraft-database/) (~5 MB).
 
 - **ReleasableAircraft_{date}.zip**  
   A daily snapshot of the FAA database, which updates at **05:30 UTC**

community/2026-02-18/ggman12_2026-02-18_5ddbb8bd.json

@@ -1,40 +0,0 @@
-[
-  {
-    "contributor_name": "JohnSmith.com",
-    "contributor_uuid": "2981c3ee-8712-5f96-84bf-732eda515a3f",
-    "creation_timestamp": "2026-02-18T22:18:11.349009+00:00",
-    "registration_number": "ZM146",
-    "tags": {
-      "citation_0": "https://assets.publishing.service.gov.uk/media/5c07a65f40f0b6705f11cf37/10389.pdf",
-      "icao_aircraft_type": "L1J",
-      "manufacturer_icao": "LOCKHEED MARTIN",
-      "manufacturer_name": "Lockheed-martin",
-      "model": "F-35B Lightning II",
-      "operator": "Royal Air Force",
-      "operator_callsign": "RAFAIR",
-      "operator_icao": "RFR",
-      "serial_number": "BK-12",
-      "type_code": "VF35"
-    },
-    "transponder_code_hex": "43C81C"
-  },
-  {
-    "contributor_name": "JohnSmith.com",
-    "contributor_uuid": "2981c3ee-8712-5f96-84bf-732eda515a3f",
-    "creation_timestamp": "2026-02-18T22:18:11.349009+00:00",
-    "registration_number": "ZM148",
-    "tags": {
-      "citation_0": "https://assets.publishing.service.gov.uk/media/5c07a65f40f0b6705f11cf37/10389.pdf",
-      "icao_aircraft_type": "L1J",
-      "manufacturer_icao": "LOCKHEED MARTIN",
-      "manufacturer_name": "Lockheed-martin",
-      "model": "F-35B Lightning II",
-      "operator": "Royal Air Force",
-      "operator_callsign": "RAFAIR",
-      "operator_icao": "RFR",
-      "serial_number": "BK-14",
-      "type_code": "VF35"
-    },
-    "transponder_code_hex": "43C811"
-  }
-]

schemas/community_submission.v1.schema.json

@@ -54,38 +54,7 @@
       "additionalProperties": {
         "$ref": "#/$defs/tagValue"
       },
-      "properties": {
-        "citation_0": {
-          "type": "string"
-        },
-        "icao_aircraft_type": {
-          "type": "string"
-        },
-        "manufacturer_icao": {
-          "type": "string"
-        },
-        "manufacturer_name": {
-          "type": "string"
-        },
-        "model": {
-          "type": "string"
-        },
-        "operator": {
-          "type": "string"
-        },
-        "operator_callsign": {
-          "type": "string"
-        },
-        "operator_icao": {
-          "type": "string"
-        },
-        "serial_number": {
-          "type": "string"
-        },
-        "type_code": {
-          "type": "string"
-        }
-      }
+      "properties": {}
     }
   },
   "allOf": [

scripts/concat_downloads.py

@@ -1,49 +0,0 @@
-#!/usr/bin/env python3
-import re
-from pathlib import Path
-import polars as pl
-
-# Find all CSV.gz files in the downloaded artifacts
-artifacts_dir = Path("downloads/adsb_artifacts")
-files = sorted(artifacts_dir.glob("*/openairframes_adsb_*.csv.gz"))
-
-if not files:
-    raise SystemExit("No CSV.gz files found in downloads/adsb_artifacts/")
-
-print(f"Found {len(files)} files to concatenate")
-
-# Extract dates from filenames to determine range
-def extract_dates(path: Path) -> tuple[str, str]:
-    """Extract start and end dates from filename"""
-    m = re.search(r"openairframes_adsb_(\d{4}-\d{2}-\d{2})_(\d{4}-\d{2}-\d{2})\.csv\.gz", path.name)
-    if m:
-        return m.group(1), m.group(2)
-    return None, None
-
-# Collect all dates
-all_dates = []
-for f in files:
-    start, end = extract_dates(f)
-    if start and end:
-        all_dates.extend([start, end])
-        print(f"  {f.name}: {start} to {end}")
-
-if not all_dates:
-    raise SystemExit("Could not extract dates from filenames")
-
-# Find earliest and latest dates
-earliest = min(all_dates)
-latest = max(all_dates)
-print(f"\nDate range: {earliest} to {latest}")
-
-# Read and concatenate all files
-print("\nReading and concatenating files...")
-frames = [pl.read_csv(f) for f in files]
-df = pl.concat(frames, how="vertical", rechunk=True)
-
-# Write output
-output_path = Path("downloads") / f"openairframes_adsb_{earliest}_{latest}.csv.gz"
-output_path.parent.mkdir(parents=True, exist_ok=True)
-df.write_csv(output_path, compression="gzip")
-
-print(f"\nWrote {output_path} with {df.height:,} rows")

scripts/download.sh

@@ -1,40 +0,0 @@
-#!/bin/bash
-
-# Create download directory
-mkdir -p downloads/adsb_artifacts
-
-# Repository from the workflow comment
-REPO="ggman12/OpenAirframes"
-
-# Get last 15 runs of the workflow and download matching artifacts
-gh run list \
-  --repo "$REPO" \
-  --workflow adsb-to-aircraft-multiple-day-run.yaml \
-  --limit 15 \
-  --json databaseId \
-  --jq '.[].databaseId' | while read -r run_id; do
-  
-  echo "Checking run ID: $run_id"
-  
-  # List artifacts for this run using the API
-  # Match pattern: openairframes_adsb-YYYY-MM-DD-YYYY-MM-DD (with second date)
-  gh api \
-    --paginate \
-    "repos/$REPO/actions/runs/$run_id/artifacts" \
-    --jq '.artifacts[] | select(.name | test("^openairframes_adsb-[0-9]{4}-[0-9]{2}-[0-9]{2}-[0-9]{4}-[0-9]{2}-[0-9]{2}$")) | .name' | while read -r artifact_name; do
-    
-    # Check if artifact directory already exists and has files
-    if [ -d "downloads/adsb_artifacts/$artifact_name" ] && [ -n "$(ls -A "downloads/adsb_artifacts/$artifact_name" 2>/dev/null)" ]; then
-      echo "  Skipping (already exists): $artifact_name"
-      continue
-    fi
-    
-    echo "  Downloading: $artifact_name"
-    gh run download "$run_id" \
-      --repo "$REPO" \
-      --name "$artifact_name" \
-      --dir "downloads/adsb_artifacts/$artifact_name"
-  done
-done
-
-echo "Download complete! Files saved to downloads/adsb_artifacts/"

scripts/download_and_concat_runs.py

@@ -1,182 +0,0 @@
-#!/usr/bin/env python3
-"""
-Download and concatenate artifacts from a specific set of workflow runs.
-
-Usage:
-    python scripts/download_and_concat_runs.py triggered_runs_20260216_123456.json
-"""
-
-import argparse
-import json
-import os
-import subprocess
-import sys
-from pathlib import Path
-
-
-def download_run_artifact(run_id, output_dir):
-    """Download artifact from a specific workflow run."""
-    print(f"  Downloading artifacts from run {run_id}...")
-    
-    cmd = [
-        'gh', 'run', 'download', str(run_id),
-        '--pattern', 'openairframes_adsb-*',
-        '--dir', output_dir
-    ]
-    
-    result = subprocess.run(cmd, capture_output=True, text=True)
-    
-    if result.returncode == 0:
-        print(f"  ✓ Downloaded")
-        return True
-    else:
-        if "no artifacts" in result.stderr.lower():
-            print(f"  ⚠ No artifacts found (workflow may still be running)")
-        else:
-            print(f"  ✗ Failed: {result.stderr}")
-        return False
-
-
-def find_csv_files(download_dir):
-    """Find all CSV.gz files in the download directory."""
-    csv_files = []
-    for root, dirs, files in os.walk(download_dir):
-        for file in files:
-            if file.endswith('.csv.gz'):
-                csv_files.append(os.path.join(root, file))
-    return sorted(csv_files)
-
-
-def concatenate_csv_files(csv_files, output_file):
-    """Concatenate CSV files in order, preserving headers."""
-    import gzip
-    
-    print(f"\nConcatenating {len(csv_files)} CSV files...")
-    
-    with gzip.open(output_file, 'wt') as outf:
-        header_written = False
-        
-        for i, csv_file in enumerate(csv_files, 1):
-            print(f"  [{i}/{len(csv_files)}] Processing {os.path.basename(csv_file)}")
-            
-            with gzip.open(csv_file, 'rt') as inf:
-                lines = inf.readlines()
-                
-                if not header_written:
-                    # Write header from first file
-                    outf.writelines(lines)
-                    header_written = True
-                else:
-                    # Skip header for subsequent files
-                    outf.writelines(lines[1:])
-    
-    print(f"\n✓ Concatenated CSV saved to: {output_file}")
-    
-    # Show file size
-    size_mb = os.path.getsize(output_file) / (1024 * 1024)
-    print(f"  Size: {size_mb:.1f} MB")
-
-
-def main():
-    parser = argparse.ArgumentParser(
-        description='Download and concatenate artifacts from workflow runs'
-    )
-    parser.add_argument(
-        'runs_file',
-        help='JSON file containing run IDs (from run_historical_adsb_action.py)'
-    )
-    parser.add_argument(
-        '--output-dir',
-        default='./downloads/historical_concat',
-        help='Directory for downloads (default: ./downloads/historical_concat)'
-    )
-    parser.add_argument(
-        '--wait',
-        action='store_true',
-        help='Wait for workflows to complete before downloading'
-    )
-    
-    args = parser.parse_args()
-    
-    # Load run IDs
-    if not os.path.exists(args.runs_file):
-        print(f"Error: File not found: {args.runs_file}")
-        sys.exit(1)
-    
-    with open(args.runs_file, 'r') as f:
-        data = json.load(f)
-    
-    runs = data['runs']
-    start_date = data['start_date']
-    end_date = data['end_date']
-    
-    print("=" * 60)
-    print("Download and Concatenate Historical Artifacts")
-    print("=" * 60)
-    print(f"Date range: {start_date} to {end_date}")
-    print(f"Workflow runs: {len(runs)}")
-    print(f"Output directory: {args.output_dir}")
-    print("=" * 60)
-    
-    # Create output directory
-    os.makedirs(args.output_dir, exist_ok=True)
-    
-    # Wait for workflows to complete if requested
-    if args.wait:
-        print("\nWaiting for workflows to complete...")
-        for run_info in runs:
-            run_id = run_info['run_id']
-            print(f"  Checking run {run_id}...")
-            
-            cmd = ['gh', 'run', 'watch', str(run_id)]
-            subprocess.run(cmd)
-    
-    # Download artifacts
-    print("\nDownloading artifacts...")
-    successful_downloads = 0
-    
-    for i, run_info in enumerate(runs, 1):
-        run_id = run_info['run_id']
-        print(f"\n[{i}/{len(runs)}] Run {run_id} ({run_info['start']} to {run_info['end']})")
-        
-        if download_run_artifact(run_id, args.output_dir):
-            successful_downloads += 1
-    
-    print(f"\n\nDownload Summary: {successful_downloads}/{len(runs)} artifacts downloaded")
-    
-    if successful_downloads == 0:
-        print("\nNo artifacts downloaded. Workflows may still be running.")
-        print("Use --wait to wait for completion, or try again later.")
-        sys.exit(1)
-    
-    # Find all CSV files
-    csv_files = find_csv_files(args.output_dir)
-    
-    if not csv_files:
-        print("\nError: No CSV files found in download directory")
-        sys.exit(1)
-    
-    print(f"\nFound {len(csv_files)} CSV file(s):")
-    for csv_file in csv_files:
-        print(f"  - {os.path.basename(csv_file)}")
-    
-    # Concatenate
-    # Calculate actual end date for filename (end_date - 1 day since it's exclusive)
-    from datetime import datetime, timedelta
-    end_dt = datetime.strptime(end_date, '%Y-%m-%d') - timedelta(days=1)
-    actual_end = end_dt.strftime('%Y-%m-%d')
-    
-    output_file = os.path.join(
-        args.output_dir,
-        f"openairframes_adsb_{start_date}_{actual_end}.csv.gz"
-    )
-    
-    concatenate_csv_files(csv_files, output_file)
-    
-    print("\n" + "=" * 60)
-    print("Done!")
-    print("=" * 60)
-
-
-if __name__ == '__main__':
-    main()

scripts/run_historical_adsb_action.py

@@ -1,215 +0,0 @@
-#!/usr/bin/env python3
-"""
-Script to trigger adsb-to-aircraft-multiple-day-run workflow runs in monthly chunks.
-
-Usage:
-    python scripts/run_historical_adsb_action.py --start-date 2025-01-01 --end-date 2025-06-01
-"""
-
-import argparse
-import subprocess
-import sys
-from datetime import datetime, timedelta
-from calendar import monthrange
-
-
-def generate_monthly_chunks(start_date_str, end_date_str):
-    """Generate date ranges in monthly chunks from start to end date.
-    
-    End dates are exclusive (e.g., to process Jan 1-31, end_date should be Feb 1).
-    """
-    start_date = datetime.strptime(start_date_str, '%Y-%m-%d')
-    end_date = datetime.strptime(end_date_str, '%Y-%m-%d')
-    
-    chunks = []
-    current = start_date
-    
-    while current < end_date:
-        # Get the first day of the next month (exclusive end)
-        _, days_in_month = monthrange(current.year, current.month)
-        month_end = current.replace(day=days_in_month)
-        next_month_start = month_end + timedelta(days=1)
-        
-        # Don't go past the global end date
-        chunk_end = min(next_month_start, end_date)
-        
-        chunks.append({
-            'start': current.strftime('%Y-%m-%d'),
-            'end': chunk_end.strftime('%Y-%m-%d')
-        })
-        
-        # Move to first day of next month
-        if next_month_start >= end_date:
-            break
-        current = next_month_start
-    
-    return chunks
-
-
-def trigger_workflow(start_date, end_date, repo='ggman12/OpenAirframes', branch='main', dry_run=False):
-    """Trigger the adsb-to-aircraft-multiple-day-run workflow via GitHub CLI."""
-    cmd = [
-        'gh', 'workflow', 'run', 'adsb-to-aircraft-multiple-day-run.yaml',
-        '--repo', repo,
-        '--ref', branch,
-        '-f', f'start_date={start_date}',
-        '-f', f'end_date={end_date}'
-    ]
-    
-    if dry_run:
-        print(f"[DRY RUN] Would run: {' '.join(cmd)}")
-        return True, None
-    
-    print(f"Triggering workflow: {start_date} to {end_date} (on {branch})")
-    result = subprocess.run(cmd, capture_output=True, text=True)
-    
-    if result.returncode == 0:
-        print(f"✓ Successfully triggered workflow for {start_date} to {end_date}")
-        
-        # Get the run ID of the workflow we just triggered
-        # Wait a moment for it to appear
-        import time
-        time.sleep(2)
-        
-        # Get the most recent run (should be the one we just triggered)
-        list_cmd = [
-            'gh', 'run', 'list',
-            '--repo', repo,
-            '--workflow', 'adsb-to-aircraft-multiple-day-run.yaml',
-            '--branch', branch,
-            '--limit', '1',
-            '--json', 'databaseId',
-            '--jq', '.[0].databaseId'
-        ]
-        list_result = subprocess.run(list_cmd, capture_output=True, text=True)
-        run_id = list_result.stdout.strip() if list_result.returncode == 0 else None
-        
-        return True, run_id
-    else:
-        print(f"✗ Failed to trigger workflow for {start_date} to {end_date}")
-        print(f"Error: {result.stderr}")
-        return False, None
-
-
-def main():
-    parser = argparse.ArgumentParser(
-        description='Trigger adsb-to-aircraft-multiple-day-run workflow runs in monthly chunks'
-    )
-    parser.add_argument(
-        '--start-date', '--start_date',
-        dest='start_date',
-        required=True,
-        help='Start date in YYYY-MM-DD format (inclusive)'
-    )
-    parser.add_argument(
-        '--end-date', '--end_date',
-        dest='end_date',
-        required=True,
-        help='End date in YYYY-MM-DD format (exclusive)'
-    )
-    parser.add_argument(
-        '--repo',
-        type=str,
-        default='ggman12/OpenAirframes',
-        help='GitHub repository (default: ggman12/OpenAirframes)'
-    )
-    parser.add_argument(
-        '--branch',
-        type=str,
-        default='main',
-        help='Branch to run the workflow on (default: main)'
-    )
-    parser.add_argument(
-        '--dry-run',
-        action='store_true',
-        help='Print commands without executing them'
-    )
-    parser.add_argument(
-        '--delay',
-        type=int,
-        default=5,
-        help='Delay in seconds between workflow triggers (default: 5)'
-    )
-    
-    args = parser.parse_args()
-    
-    # Validate dates
-    try:
-        start = datetime.strptime(args.start_date, '%Y-%m-%d')
-        end = datetime.strptime(args.end_date, '%Y-%m-%d')
-        if start > end:
-            print("Error: start_date must be before or equal to end_date")
-            sys.exit(1)
-    except ValueError as e:
-        print(f"Error: Invalid date format - {e}")
-        sys.exit(1)
-    
-    # Generate monthly chunks
-    chunks = generate_monthly_chunks(args.start_date, args.end_date)
-    
-    print(f"\nGenerating {len(chunks)} monthly workflow runs on branch '{args.branch}' (repo: {args.repo}):")
-    for i, chunk in enumerate(chunks, 1):
-        print(f"  {i}. {chunk['start']} to {chunk['end']}")
-    
-    if not args.dry_run:
-        response = input(f"\nProceed with triggering {len(chunks)} workflows on '{args.branch}'? [y/N]: ")
-        if response.lower() != 'y':
-            print("Cancelled.")
-            sys.exit(0)
-    
-    print()
-    
-    # Trigger workflows
-    import time
-    success_count = 0
-    triggered_runs = []
-    
-    for i, chunk in enumerate(chunks, 1):
-        print(f"\n[{i}/{len(chunks)}] ", end='')
-        
-        success, run_id = trigger_workflow(
-            chunk['start'],
-            chunk['end'],
-            repo=args.repo,
-            branch=args.branch,
-            dry_run=args.dry_run
-        )
-        
-        if success:
-            success_count += 1
-            if run_id:
-                triggered_runs.append({
-                    'run_id': run_id,
-                    'start': chunk['start'],
-                    'end': chunk['end']
-                })
-        
-        # Add delay between triggers (except for last one)
-        if i < len(chunks) and not args.dry_run:
-            time.sleep(args.delay)
-    
-    print(f"\n\nSummary: {success_count}/{len(chunks)} workflows triggered successfully")
-    
-    # Save triggered run IDs to a file
-    if triggered_runs and not args.dry_run:
-        import json
-        timestamp = datetime.now().strftime('%Y%m%d_%H%M%S')
-        runs_file = f"./output/triggered_runs_{timestamp}.json"
-        with open(runs_file, 'w') as f:
-            json.dump({
-                'start_date': args.start_date,
-                'end_date': args.end_date,
-                'repo': args.repo,
-                'branch': args.branch,
-                'runs': triggered_runs
-            }, f, indent=2)
-        print(f"\nRun IDs saved to: {runs_file}")
-        print(f"\nTo download and concatenate these artifacts, run:")
-        print(f"  python scripts/download_and_concat_runs.py {runs_file}")
-    
-    if success_count < len(chunks):
-        sys.exit(1)
-
-
-if __name__ == '__main__':
-    main()

scripts/run_main_isolated.py

@@ -1,82 +0,0 @@
-#!/usr/bin/env python3
-"""
-Run src.adsb.main in an isolated git worktree so edits in the main
-working tree won't affect subprocess imports during the run.
-
-Usage:
-    python scripts/run_main_isolated.py 2026-01-01
-    python scripts/run_main_isolated.py --start_date 2026-01-01 --end_date 2026-01-03
-"""
-import argparse
-import os
-import shutil
-import subprocess
-import sys
-from datetime import datetime, timezone
-from pathlib import Path
-
-
-def run(
-    cmd: list[str],
-    *,
-    cwd: Path | None = None,
-    check: bool = True,
-) -> subprocess.CompletedProcess:
-    print(f"\n>>> {' '.join(cmd)}")
-    return subprocess.run(cmd, cwd=cwd, check=check)
-
-
-def main() -> int:
-    parser = argparse.ArgumentParser(description="Run src.adsb.main in an isolated worktree")
-    parser.add_argument("date", nargs="?", help="Single date to process (YYYY-MM-DD)")
-    parser.add_argument("--start_date", help="Start date (inclusive, YYYY-MM-DD)")
-    parser.add_argument("--end_date", help="End date (exclusive, YYYY-MM-DD)")
-    parser.add_argument("--concat_with_latest_csv", action="store_true", help="Also concatenate with latest CSV from GitHub releases")
-    args = parser.parse_args()
-
-    if args.date and (args.start_date or args.end_date):
-        raise SystemExit("Use a single date or --start_date/--end_date, not both.")
-
-    if args.date:
-        datetime.strptime(args.date, "%Y-%m-%d")
-        main_args = ["--date", args.date]
-    else:
-        if not args.start_date or not args.end_date:
-            raise SystemExit("Provide --start_date and --end_date, or a single date.")
-        datetime.strptime(args.start_date, "%Y-%m-%d")
-        datetime.strptime(args.end_date, "%Y-%m-%d")
-        main_args = ["--start_date", args.start_date, "--end_date", args.end_date]
-
-    if args.concat_with_latest_csv:
-        main_args.append("--concat_with_latest_csv")
-
-    repo_root = Path(__file__).resolve().parents[1]
-    snapshots_root = repo_root / ".snapshots"
-    snapshots_root.mkdir(exist_ok=True)
-
-    timestamp = datetime.now(timezone.utc).strftime("%Y%m%d_%H%M%S")
-    snapshot_root = snapshots_root / f"run_{timestamp}"
-    snapshot_src = snapshot_root / "src"
-
-    exit_code = 0
-    try:
-        shutil.copytree(repo_root / "src", snapshot_src)
-
-        runner = (
-            "import sys, runpy; "
-            f"sys.path.insert(0, {repr(str(snapshot_root))}); "
-            f"sys.argv = ['src.adsb.main'] + {main_args!r}; "
-            "runpy.run_module('src.adsb.main', run_name='__main__')"
-        )
-        cmd = [sys.executable, "-c", runner]
-        run(cmd, cwd=repo_root)
-    except subprocess.CalledProcessError as exc:
-        exit_code = exc.returncode
-    finally:
-        shutil.rmtree(snapshot_root, ignore_errors=True)
-
-    return exit_code
-
-
-if __name__ == "__main__":
-    raise SystemExit(main())

scripts/scrape_theairtraffic.py

@@ -1,242 +0,0 @@
-#!/usr/bin/env python3
-"""
-Parse TheAirTraffic Database CSV and produce community_submission.v1 JSON.
-
-Source: "TheAirTraffic Database - Aircraft 2.csv"
-Output: community/YYYY-MM-DD/theairtraffic_<date>_<hash>.json
-
-Categories in the spreadsheet columns (paired: name, registrations, separator):
-  Col  1-3:  Business
-  Col  4-6:  Government
-  Col  7-9:  People
-  Col 10-12: Sports
-  Col 13-15: Celebrity
-  Col 16-18: State Govt./Law
-  Col 19-21: Other
-  Col 22-24: Test Aircraft
-  Col 25-27: YouTubers
-  Col 28-30: Formula 1 VIP's
-  Col 31-33: Active GII's and GIII's  (test/demo aircraft)
-  Col 34-37: Russia & Ukraine          (extra col for old/new)
-  Col 38-40: Helicopters & Blimps
-  Col 41-43: Unique Reg's
-  Col 44-46: Saudi & UAE
-  Col 47-49: Schools
-  Col 50-52: Special Charter
-  Col 53-55: Unknown Owners
-  Col 56-59: Frequent Flyers           (extra cols: name, aircraft, logged, hours)
-"""
-
-import csv
-import json
-import hashlib
-import re
-import sys
-import uuid
-from datetime import datetime, timezone
-from pathlib import Path
-
-# ── Category mapping ────────────────────────────────────────────────────────
-# Each entry: (name_col, reg_col, owner_category_tags)
-# owner_category_tags is a dict of tag keys to add beyond "owner"
-CATEGORY_COLUMNS = [
-    # (name_col, reg_col, {tag_key: tag_value, ...})
-    (1,  2,  {"owner_category_0": "business"}),
-    (4,  5,  {"owner_category_0": "government"}),
-    (7,  8,  {"owner_category_0": "celebrity"}),
-    (10, 11, {"owner_category_0": "sports"}),
-    (13, 14, {"owner_category_0": "celebrity"}),
-    (16, 17, {"owner_category_0": "government", "owner_category_1": "law_enforcement"}),
-    (19, 20, {"owner_category_0": "other"}),
-    (22, 23, {"owner_category_0": "test_aircraft"}),
-    (25, 26, {"owner_category_0": "youtuber", "owner_category_1": "celebrity"}),
-    (28, 29, {"owner_category_0": "celebrity", "owner_category_1": "motorsport"}),
-    (31, 32, {"owner_category_0": "test_aircraft"}),
-    # Russia & Ukraine: col 34=name, col 35 or 36 may have reg
-    (34, 35, {"owner_category_0": "russia_ukraine"}),
-    (38, 39, {"owner_category_0": "celebrity", "category": "helicopter_or_blimp"}),
-    (41, 42, {"owner_category_0": "other"}),
-    (44, 45, {"owner_category_0": "government", "owner_category_1": "royal_family"}),
-    (47, 48, {"owner_category_0": "education"}),
-    (50, 51, {"owner_category_0": "charter"}),
-    (53, 54, {"owner_category_0": "unknown"}),
-    (56, 57, {"owner_category_0": "celebrity"}),   # Frequent Flyers name col, aircraft col
-]
-
-# First data row index (0-based) in the CSV
-DATA_START_ROW = 4
-
-# ── Contributor info ────────────────────────────────────────────────────────
-CONTRIBUTOR_NAME = "TheAirTraffic"
-# Deterministic UUID v5 from contributor name
-CONTRIBUTOR_UUID = str(uuid.uuid5(uuid.NAMESPACE_URL, "https://theairtraffic.com"))
-
-# Citation
-CITATION = "https://docs.google.com/spreadsheets/d/1JHhfJBnJPNBA6TgiSHjkXFkHBdVTTz_nXxaUDRWcHpk"
-
-
-def looks_like_military_serial(reg: str) -> bool:
-    """
-    Detect military-style serials like 92-9000, 82-8000, 98-0001
-    or pure numeric IDs like 929000, 828000, 980001.
-    These aren't standard civil registrations; use openairframes_id.
-    """
-    # Pattern: NN-NNNN
-    if re.match(r'^\d{2}-\d{4}$', reg):
-        return True
-    # Pure 6-digit numbers (likely ICAO hex or military mode-S)
-    if re.match(r'^\d{6}$', reg):
-        return True
-    # Short numeric-only (1-5 digits) like "01", "02", "676"
-    if re.match(r'^\d{1,5}$', reg):
-        return True
-    return False
-
-
-def normalize_reg(raw: str) -> str:
-    """Clean up a registration string."""
-    reg = raw.strip().rstrip(',').strip()
-    # Remove carriage returns and other whitespace
-    reg = reg.replace('\r', '').replace('\n', '').strip()
-    return reg
-
-
-def parse_regs(cell_value: str) -> list[str]:
-    """
-    Parse a cell that may contain one or many registrations,
-    separated by commas, possibly wrapped in quotes.
-    """
-    if not cell_value or not cell_value.strip():
-        return []
-
-    # Some cells have ADS-B exchange URLs – skip those
-    if 'globe.adsbexchange.com' in cell_value:
-        return []
-    if cell_value.strip() in ('.', ',', ''):
-        return []
-
-    results = []
-    # Split on comma
-    parts = cell_value.split(',')
-    for part in parts:
-        reg = normalize_reg(part)
-        if not reg:
-            continue
-        # Skip URLs, section labels, etc.
-        if reg.startswith('http') or reg.startswith('Link') or reg == 'Section 1':
-            continue
-        # Skip if it's just whitespace or dots
-        if reg in ('.', '..', '...'):
-            continue
-        results.append(reg)
-    return results
-
-
-def make_submission(
-    reg: str,
-    owner: str,
-    category_tags: dict[str, str],
-) -> dict:
-    """Build a single community_submission.v1 object."""
-
-    entry: dict = {}
-
-    # Decide identifier field
-    if looks_like_military_serial(reg):
-        entry["openairframes_id"] = reg
-    else:
-        entry["registration_number"] = reg
-
-    # Tags
-    tags: dict = {
-        "citation_0": CITATION,
-    }
-    if owner:
-        tags["owner"] = owner.strip()
-    tags.update(category_tags)
-    entry["tags"] = tags
-
-    return entry
-
-
-def main():
-    csv_path = Path(sys.argv[1]) if len(sys.argv) > 1 else Path(
-        "/Users/jonahgoode/Downloads/TheAirTraffic Database - Aircraft 2.csv"
-    )
-
-    if not csv_path.exists():
-        print(f"ERROR: CSV not found at {csv_path}", file=sys.stderr)
-        sys.exit(1)
-
-    # Read CSV
-    with open(csv_path, 'r', encoding='utf-8-sig') as f:
-        reader = csv.reader(f)
-        rows = list(reader)
-
-    print(f"Read {len(rows)} rows from {csv_path.name}")
-
-    date_str = datetime.now(timezone.utc).strftime("%Y-%m-%d")
-
-    submissions: list[dict] = []
-    seen: set[tuple] = set()  # (reg, owner) dedup
-
-    for row_idx in range(DATA_START_ROW, len(rows)):
-        row = rows[row_idx]
-        if len(row) < 3:
-            continue
-
-        for name_col, reg_col, cat_tags in CATEGORY_COLUMNS:
-            if reg_col >= len(row) or name_col >= len(row):
-                continue
-
-            owner_raw = row[name_col].strip().rstrip(',').strip()
-            reg_raw = row[reg_col]
-
-            # Clean owner name
-            owner = owner_raw.replace('\r', '').replace('\n', '').strip()
-            if not owner or owner in ('.', ',', 'Section 1'):
-                continue
-            # Skip header-like values
-            if owner.startswith('http') or owner.startswith('Link '):
-                continue
-
-            regs = parse_regs(reg_raw)
-            if not regs:
-                # For Russia & Ukraine, try the next column too (col 35 might have old reg, col 36 new)
-                if name_col == 34 and reg_col + 1 < len(row):
-                    regs = parse_regs(row[reg_col + 1])
-
-            for reg in regs:
-                key = (reg, owner)
-                if key in seen:
-                    continue
-                seen.add(key)
-                submissions.append(make_submission(reg, owner, cat_tags))
-
-    print(f"Generated {len(submissions)} submissions")
-
-    # Write output
-    proj_root = Path(__file__).resolve().parent.parent
-    out_dir = proj_root / "community" / date_str
-    out_dir.mkdir(parents=True, exist_ok=True)
-
-    out_file = out_dir / f"theairtraffic_{date_str}.json"
-
-    with open(out_file, 'w', encoding='utf-8') as f:
-        json.dump(submissions, f, indent=2, ensure_ascii=False)
-
-    print(f"Written to {out_file}")
-    print(f"Sample entry:\n{json.dumps(submissions[0], indent=2)}")
-
-    # Quick stats
-    cats = {}
-    for s in submissions:
-        c = s['tags'].get('owner_category_0', 'NONE')
-        cats[c] = cats.get(c, 0) + 1
-    print("\nCategory breakdown:")
-    for c, n in sorted(cats.items(), key=lambda x: -x[1]):
-        print(f"  {c}: {n}")
-
-
-if __name__ == "__main__":
-    main()

scripts/validate_theairtraffic.py

@@ -1,69 +0,0 @@
-#!/usr/bin/env python3
-"""Validate the generated theairtraffic JSON output."""
-import json
-import glob
-import sys
-
-# Find the latest output
-files = sorted(glob.glob("community/2026-02-*/theairtraffic_*.json"))
-if not files:
-    print("No output files found!")
-    sys.exit(1)
-
-path = files[-1]
-print(f"Validating: {path}")
-
-with open(path) as f:
-    data = json.load(f)
-
-print(f"Total entries: {len(data)}")
-
-# Check military serial handling
-mil = [d for d in data if "openairframes_id" in d]
-print(f"\nEntries using openairframes_id: {len(mil)}")
-for m in mil[:10]:
-    print(f"  {m['openairframes_id']} -> owner: {m['tags'].get('owner','?')}")
-
-# Check youtuber entries
-yt = [d for d in data if d["tags"].get("owner_category_0") == "youtuber"]
-print(f"\nYouTuber entries: {len(yt)}")
-for y in yt[:5]:
-    reg = y.get("registration_number", y.get("openairframes_id"))
-    c0 = y["tags"].get("owner_category_0")
-    c1 = y["tags"].get("owner_category_1")
-    print(f"  {reg} -> owner: {y['tags']['owner']}, cat0: {c0}, cat1: {c1}")
-
-# Check US Govt / military
-gov = [d for d in data if d["tags"].get("owner") == "United States of America 747/757"]
-print(f"\nUSA 747/757 entries: {len(gov)}")
-for g in gov:
-    oid = g.get("openairframes_id", g.get("registration_number"))
-    print(f"  {oid}")
-
-# Schema validation
-issues = 0
-for i, d in enumerate(data):
-    has_id = any(k in d for k in ["registration_number", "transponder_code_hex", "openairframes_id"])
-    if not has_id:
-        print(f"  Entry {i}: no identifier!")
-        issues += 1
-    if "tags" not in d:
-        print(f"  Entry {i}: no tags!")
-        issues += 1
-    # Check tag key format
-    for k in d.get("tags", {}):
-        import re
-        if not re.match(r"^[a-z][a-z0-9_]{0,63}$", k):
-            print(f"  Entry {i}: invalid tag key '{k}'")
-            issues += 1
-
-print(f"\nSchema issues: {issues}")
-
-# Category breakdown
-cats = {}
-for s in data:
-    c = s["tags"].get("owner_category_0", "NONE")
-    cats[c] = cats.get(c, 0) + 1
-print("\nCategory breakdown:")
-for c, n in sorted(cats.items(), key=lambda x: -x[1]):
-    print(f"  {c}: {n}")

src/adsb/combine_chunks_to_csv.py

@@ -0,0 +1,261 @@
+"""
+Combines chunk parquet files and compresses to final aircraft CSV.
+This is the reduce phase of the map-reduce pipeline.
+
+Supports both single-day (daily) and multi-day (historical) modes.
+
+Memory-efficient: processes each chunk separately, compresses, then combines.
+
+Usage:
+    # Daily mode
+    python -m src.adsb.combine_chunks_to_csv --chunks-dir data/output/adsb_chunks
+    
+    # Historical mode
+    python -m src.adsb.combine_chunks_to_csv --chunks-dir data/output/adsb_chunks --start-date 2024-01-01 --end-date 2024-01-07 --skip-base
+"""
+import gc
+import gzip
+import os
+import sys
+import glob
+import argparse
+from datetime import datetime, timedelta, timezone
+
+import polars as pl
+
+from src.adsb.download_adsb_data_to_parquet import OUTPUT_DIR, get_resource_usage
+from src.adsb.compress_adsb_to_aircraft_data import compress_multi_icao_df, COLUMNS
+
+
+DEFAULT_CHUNK_DIR = os.path.join(OUTPUT_DIR, "adsb_chunks")
+FINAL_OUTPUT_DIR = "./data/openairframes"
+os.makedirs(FINAL_OUTPUT_DIR, exist_ok=True)
+
+
+def get_target_day() -> datetime:
+    """Get yesterday's date (the day we're processing)."""
+    return datetime.now(timezone.utc) - timedelta(days=1)
+
+
+def process_single_chunk(chunk_path: str, delete_after_load: bool = False) -> pl.DataFrame:
+    """Load and compress a single chunk parquet file.
+    
+    Args:
+        chunk_path: Path to parquet file
+        delete_after_load: If True, delete the parquet file after loading to free disk space
+    """
+    print(f"Processing {os.path.basename(chunk_path)}... | {get_resource_usage()}")
+    
+    # Load chunk - only columns we need
+    needed_columns = ['time', 'icao'] + COLUMNS
+    df = pl.read_parquet(chunk_path, columns=needed_columns)
+    print(f"  Loaded {len(df)} rows")
+    
+    # Delete file immediately after loading to free disk space
+    if delete_after_load:
+        try:
+            os.remove(chunk_path)
+            print(f"  Deleted {chunk_path} to free disk space")
+        except Exception as e:
+            print(f"  Warning: Failed to delete {chunk_path}: {e}")
+    
+    # Compress to aircraft records (one per ICAO) using shared function
+    compressed = compress_multi_icao_df(df, verbose=True)
+    print(f"  Compressed to {len(compressed)} aircraft records")
+    
+    del df
+    gc.collect()
+    
+    return compressed
+
+
+def combine_compressed_chunks(compressed_dfs: list[pl.DataFrame]) -> pl.DataFrame:
+    """Combine multiple compressed DataFrames.
+    
+    Since chunks are partitioned by ICAO hash, each ICAO only appears in one chunk.
+    No deduplication needed here - just concatenate.
+    """
+    print(f"Combining {len(compressed_dfs)} compressed chunks... | {get_resource_usage()}")
+    
+    # Concat all
+    combined = pl.concat(compressed_dfs)
+    print(f"Combined: {len(combined)} records")
+    
+    return combined
+
+
+def download_and_merge_base_release(compressed_df: pl.DataFrame) -> tuple[pl.DataFrame, str | None]:
+    """Download base release and merge with new data.
+    
+    Returns:
+        Tuple of (merged_df, earliest_date_str) where earliest_date_str is None if no base release was merged
+    """
+    from src.get_latest_release import download_latest_aircraft_adsb_csv
+    
+    print("Downloading base ADS-B release...")
+    base_path = download_latest_aircraft_adsb_csv(
+        output_dir="./data/openairframes_base"
+    )
+    print(f"Download returned: {base_path}")
+    
+    print(f"Loading base release from {base_path}")
+    
+    # Extract start date from filename (e.g., openairframes_adsb_2025-05-01_2026-02-14.csv.gz)
+    import re
+    filename = os.path.basename(str(base_path))
+    match = re.search(r'openairframes_adsb_(\d{4}-\d{2}-\d{2})_', filename)
+    earliest_date = match.group(1) if match else None
+    print(f"Start date from base filename: {earliest_date}")
+    
+    # Read CSV with schema matching the new data
+    base_df = pl.read_csv(base_path, schema=compressed_df.schema)
+    print(f"Base release has {len(base_df)} records")
+    
+    # Ensure columns match
+    base_cols = set(base_df.columns)
+    new_cols = set(compressed_df.columns)
+    print(f"Base columns: {sorted(base_cols)}")
+    print(f"New columns: {sorted(new_cols)}")
+    
+    # Add missing columns
+    for col in new_cols - base_cols:
+        base_df = base_df.with_columns(pl.lit(None).alias(col))
+    for col in base_cols - new_cols:
+        compressed_df = compressed_df.with_columns(pl.lit(None).alias(col))
+    
+    # Reorder columns to match
+    compressed_df = compressed_df.select(base_df.columns)
+    
+    # Concat and deduplicate by icao (keep new data - it comes last)
+    combined = pl.concat([base_df, compressed_df])
+    print(f"After concat: {len(combined)} records")
+
+    return combined, earliest_date
+
+def cleanup_chunks(output_id: str, chunks_dir: str):
+    """Delete chunk parquet files after successful merge."""
+    pattern = os.path.join(chunks_dir, f"chunk_*_{output_id}.parquet")
+    chunk_files = glob.glob(pattern)
+    for f in chunk_files:
+        try:
+            os.remove(f)
+            print(f"Deleted {f}")
+        except Exception as e:
+            print(f"Failed to delete {f}: {e}")
+
+
+def find_chunk_files(chunks_dir: str, output_id: str) -> list[str]:
+    """Find chunk parquet files matching the output ID."""
+    pattern = os.path.join(chunks_dir, f"chunk_*_{output_id}.parquet")
+    chunk_files = sorted(glob.glob(pattern))
+    
+    if not chunk_files:
+        # Try recursive search for historical mode with merged artifacts
+        pattern = os.path.join(chunks_dir, "**", "*.parquet")
+        chunk_files = sorted(glob.glob(pattern, recursive=True))
+    
+    return chunk_files
+
+
+def main():
+    parser = argparse.ArgumentParser(description="Combine chunk parquets to final CSV")
+    parser.add_argument("--date", type=str, help="Single date in YYYY-MM-DD format (default: yesterday)")
+    parser.add_argument("--start-date", type=str, help="Start date for range (YYYY-MM-DD)")
+    parser.add_argument("--end-date", type=str, help="End date for range (YYYY-MM-DD)")
+    parser.add_argument("--chunks-dir", type=str, default=DEFAULT_CHUNK_DIR, help="Directory containing chunk parquet files")
+    parser.add_argument("--skip-base", action="store_true", help="Skip downloading and merging base release")
+    parser.add_argument("--keep-chunks", action="store_true", help="Keep chunk files after merging")
+    parser.add_argument("--stream", action="store_true", help="Delete parquet files immediately after loading to save disk space")
+    args = parser.parse_args()
+    
+    # Determine output ID and filename based on mode
+    if args.start_date and args.end_date:
+        # Historical mode
+        output_id = f"{args.start_date}_{args.end_date}"
+        output_filename = f"openairframes_adsb_{args.start_date}_{args.end_date}.csv.gz"
+        print(f"Combining chunks for date range: {args.start_date} to {args.end_date}")
+    else:
+        # Daily mode - use same date for start and end
+        if args.date:
+            target_day = datetime.strptime(args.date, "%Y-%m-%d")
+        else:
+            target_day = get_target_day()
+        
+        date_str = target_day.strftime("%Y-%m-%d")
+        output_id = date_str
+        output_filename = f"openairframes_adsb_{date_str}_{date_str}.csv.gz"
+        print(f"Combining chunks for {date_str}")
+    
+    chunks_dir = args.chunks_dir
+    print(f"Chunks directory: {chunks_dir}")
+    print(f"Resource usage at start: {get_resource_usage()}")
+    
+    # Find chunk files
+    chunk_files = find_chunk_files(chunks_dir, output_id)
+    
+    if not chunk_files:
+        print(f"No chunk files found in: {chunks_dir}")
+        sys.exit(1)
+    
+    print(f"Found {len(chunk_files)} chunk files")
+    
+    # Process each chunk separately to save memory
+    # With --stream, delete parquet files immediately after loading to save disk space
+    compressed_chunks = []
+    for chunk_path in chunk_files:
+        compressed = process_single_chunk(chunk_path, delete_after_load=args.stream)
+        compressed_chunks.append(compressed)
+        gc.collect()
+    
+    # Combine all compressed chunks
+    combined = combine_compressed_chunks(compressed_chunks)
+    
+    # Free memory from individual chunks
+    del compressed_chunks
+    gc.collect()
+    print(f"After combining: {get_resource_usage()}")
+    
+    # Merge with base release (unless skipped)
+    base_start_date = None
+    if not args.skip_base:
+        combined, base_start_date = download_and_merge_base_release(combined)
+        
+        # Update filename if we merged with base release and got a start date
+        if base_start_date and not (args.start_date and args.end_date):
+            # Only update filename for daily mode when base was merged
+            output_filename = f"openairframes_adsb_{base_start_date}_{date_str}.csv.gz"
+            print(f"Updated filename to reflect date range: {output_filename}")
+    
+    # Convert list columns to strings for CSV compatibility
+    for col in combined.columns:
+        if combined[col].dtype == pl.List:
+            combined = combined.with_columns(
+                pl.col(col).list.join(",").alias(col)
+            )
+    
+    # Sort by time for consistent output
+    if 'time' in combined.columns:
+        combined = combined.sort('time')
+    
+    # Replace empty strings with null across all string columns to avoid quoted empty strings
+    for col in combined.columns:
+        if combined[col].dtype == pl.Utf8:
+            combined = combined.with_columns(
+                pl.when(pl.col(col) == "").then(None).otherwise(pl.col(col)).alias(col)
+            )
+    
+    # Write final CSV
+    output_path = os.path.join(FINAL_OUTPUT_DIR, output_filename)
+    with gzip.open(output_path, "wb", compresslevel=9) as f:
+        combined.write_csv(f, null_value='', quote_style='necessary')
+    print(f"Wrote {len(combined)} records to {output_path}")
+    
+    # Cleanup
+    if not args.keep_chunks:
+        cleanup_chunks(output_id, chunks_dir)
+    
+    print(f"Done! | {get_resource_usage()}")
+
+
+if __name__ == "__main__":
+    main()

src/adsb/compress_adsb_to_aircraft_data.py

@@ -5,6 +5,23 @@ import polars as pl
 COLUMNS = ['dbFlags', 'ownOp', 'year', 'desc', 'aircraft_category', 'r', 't']
 
 
+def deduplicate_by_signature(df: pl.DataFrame) -> pl.DataFrame:
+    """For each icao, keep only the earliest row with each unique signature.
+    
+    This is used for deduplicating across multiple compressed chunks.
+    """
+    # Create signature column
+    df = df.with_columns(
+        pl.concat_str([pl.col(c).cast(pl.Utf8).fill_null("") for c in COLUMNS], separator="|").alias("_signature")
+    )
+    # Group by icao and signature, take first row (earliest due to time sort)
+    df = df.sort("time")
+    df_deduped = df.group_by(["icao", "_signature"]).first()
+    df_deduped = df_deduped.drop("_signature")
+    df_deduped = df_deduped.sort("time")
+    return df_deduped
+
+
 def compress_df_polars(df: pl.DataFrame, icao: str) -> pl.DataFrame:
     """Compress a single ICAO group to its most informative row using Polars."""
     # Create signature string
@@ -82,6 +99,9 @@ def compress_df_polars(df: pl.DataFrame, icao: str) -> pl.DataFrame:
 def compress_multi_icao_df(df: pl.DataFrame, verbose: bool = True) -> pl.DataFrame:
     """Compress a DataFrame with multiple ICAOs to one row per ICAO.
     
+    This is the main entry point for compressing ADS-B data.
+    Used by both daily GitHub Actions runs and historical AWS runs.
+    
     Args:
         df: DataFrame with columns ['time', 'icao'] + COLUMNS
         verbose: Whether to print progress
@@ -100,27 +120,29 @@ def compress_multi_icao_df(df: pl.DataFrame, verbose: bool = True) -> pl.DataFra
         if col in df.columns:
             df = df.with_columns(pl.col(col).cast(pl.Utf8).fill_null(""))
     
-    # Quick deduplication of exact duplicates
+    # First pass: quick deduplication of exact duplicates
     df = df.unique(subset=['icao'] + COLUMNS, keep='first')
     if verbose:
         print(f"After quick dedup: {df.height} records")
     
-    # Compress per ICAO
+    # Second pass: sophisticated compression per ICAO
     if verbose:
         print("Compressing per ICAO...")
     
+    # Process each ICAO group
     icao_groups = df.partition_by('icao', as_dict=True, maintain_order=True)
     compressed_dfs = []
     
     for icao_key, group_df in icao_groups.items():
-        icao = icao_key[0]
+        # partition_by with as_dict=True returns tuple keys, extract first element
+        icao = icao_key[0] if isinstance(icao_key, tuple) else icao_key
         compressed = compress_df_polars(group_df, str(icao))
         compressed_dfs.append(compressed)
     
     if compressed_dfs:
         df_compressed = pl.concat(compressed_dfs)
     else:
-        df_compressed = df.head(0)
+        df_compressed = df.head(0)  # Empty with same schema
     
     if verbose:
         print(f"After compress: {df_compressed.height} records")
@@ -133,22 +155,45 @@ def compress_multi_icao_df(df: pl.DataFrame, verbose: bool = True) -> pl.DataFra
     return df_compressed
 
 
-def load_parquet_part(part_id: int, date: str) -> pl.DataFrame:
-    """Load a single parquet part file for a date.
-    
-    Args:
-        part_id: Part ID (e.g., 1, 2, 3)
-        date: Date string in YYYY-MM-DD format
-    
-    Returns:
-        DataFrame with ADS-B data
-    """
+def load_raw_adsb_for_day(day):
+    """Load raw ADS-B data for a day from parquet file."""
+    from datetime import timedelta
     from pathlib import Path
     
-    parquet_file = Path(f"data/output/parquet_output/part_{part_id}_{date}.parquet")
+    start_time = day.replace(hour=0, minute=0, second=0, microsecond=0)
+    
+    # Check for parquet file first
+    version_date = f"v{start_time.strftime('%Y.%m.%d')}"
+    parquet_file = Path(f"data/output/parquet_output/{version_date}.parquet")
     
     if not parquet_file.exists():
-        print(f"Parquet file not found: {parquet_file}")
+        # Try to generate parquet file by calling the download function
+        print(f"  Parquet file not found: {parquet_file}")
+        print(f"  Attempting to download and generate parquet for {start_time.strftime('%Y-%m-%d')}...")
+        
+        from download_adsb_data_to_parquet import create_parquet_for_day
+        result_path = create_parquet_for_day(start_time, keep_folders=False)
+        
+        if result_path:
+            print(f"  Successfully generated parquet file: {result_path}")
+        else:
+            raise Exception("Failed to generate parquet file")
+    
+    if parquet_file.exists():
+        print(f"  Loading from parquet: {parquet_file}")
+        df = pl.read_parquet(
+            parquet_file, 
+            columns=['time', 'icao', 'r', 't', 'dbFlags', 'ownOp', 'year', 'desc', 'aircraft_category']
+        )
+        
+        # Convert to timezone-naive datetime
+        if df["time"].dtype == pl.Datetime:
+            df = df.with_columns(pl.col("time").dt.replace_time_zone(None))
+        
+        return df
+    else:
+        # Return empty DataFrame if parquet file doesn't exist
+        print(f"  No data available for {start_time.strftime('%Y-%m-%d')}")
         return pl.DataFrame(schema={
             'time': pl.Datetime,
             'icao': pl.Utf8,
@@ -160,33 +205,17 @@ def load_parquet_part(part_id: int, date: str) -> pl.DataFrame:
             'desc': pl.Utf8,
             'aircraft_category': pl.Utf8
         })
-    
-    print(f"Loading from parquet: {parquet_file}")
-    df = pl.read_parquet(
-        parquet_file,
-        columns=['time', 'icao', 'r', 't', 'dbFlags', 'ownOp', 'year', 'desc', 'aircraft_category']
-    )
-    
-    # Convert to timezone-naive datetime
-    if df["time"].dtype == pl.Datetime:
-        df = df.with_columns(pl.col("time").dt.replace_time_zone(None))
-    os.remove(parquet_file)
-    return df
 
 
-def compress_parquet_part(part_id: int, date: str) -> pl.DataFrame:
-    """Load and compress a single parquet part file."""
-    df = load_parquet_part(part_id, date)
-    
+def load_historical_for_day(day):
+    """Load and compress historical ADS-B data for a day."""
+    df = load_raw_adsb_for_day(day)
     if df.height == 0:
         return df
-
-    # Filter to rows within the given date (UTC-naive). This is because sometimes adsb.lol export can have rows at 00:00:00 of next day or similar.
-    date_lit = pl.lit(date).str.strptime(pl.Date, "%Y-%m-%d")
-    df = df.filter(pl.col("time").dt.date() == date_lit)
     
-    print(f"Loaded {df.height} raw records for part {part_id}, date {date}")
+    print(f"Loaded {df.height} raw records for {day.strftime('%Y-%m-%d')}")
     
+    # Use shared compression function
     return compress_multi_icao_df(df, verbose=True)
 
 
@@ -194,4 +223,52 @@ def concat_compressed_dfs(df_base, df_new):
     """Concatenate base and new compressed dataframes, keeping the most informative row per ICAO."""
     # Combine both dataframes
     df_combined = pl.concat([df_base, df_new])
-    return df_combined
+    
+    # Sort by ICAO and time
+    df_combined = df_combined.sort(['icao', 'time'])
+    
+    # Fill null values
+    for col in COLUMNS:
+        if col in df_combined.columns:
+            df_combined = df_combined.with_columns(pl.col(col).fill_null(""))
+    
+    # Apply compression logic per ICAO to get the best row
+    icao_groups = df_combined.partition_by('icao', as_dict=True, maintain_order=True)
+    compressed_dfs = []
+    
+    for icao, group_df in icao_groups.items():
+        compressed = compress_df_polars(group_df, icao)
+        compressed_dfs.append(compressed)
+    
+    if compressed_dfs:
+        df_compressed = pl.concat(compressed_dfs)
+    else:
+        df_compressed = df_combined.head(0)
+    
+    # Sort by time
+    df_compressed = df_compressed.sort('time')
+    
+    return df_compressed
+
+
+def get_latest_aircraft_adsb_csv_df():
+    """Download and load the latest ADS-B CSV from GitHub releases."""
+    from get_latest_release import download_latest_aircraft_adsb_csv
+    import re
+    
+    csv_path = download_latest_aircraft_adsb_csv()
+    df = pl.read_csv(csv_path, null_values=[""])
+    
+    # Fill nulls with empty strings
+    for col in df.columns:
+        if df[col].dtype == pl.Utf8:
+            df = df.with_columns(pl.col(col).fill_null(""))
+    
+    # Extract start date from filename pattern: openairframes_adsb_{start_date}_{end_date}.csv[.gz]
+    match = re.search(r"openairframes_adsb_(\d{4}-\d{2}-\d{2})_", str(csv_path))
+    if not match:
+        raise ValueError(f"Could not extract date from filename: {csv_path.name}")
+    
+    date_str = match.group(1)
+    return df, date_str
+

src/adsb/concat_parquet_to_final.py

@@ -1,67 +0,0 @@
-from pathlib import Path
-import polars as pl
-import argparse
-import os
-OUTPUT_DIR = Path("./data/output")
-CORRECT_ORDER_OF_COLUMNS = ["time", "icao", "r", "t", "dbFlags", "ownOp", "year", "desc", "aircraft_category"]
-
-def main():
-    parser = argparse.ArgumentParser(description="Concatenate compressed parquet files for a single day")
-    parser.add_argument("--date", type=str, required=True, help="Date in YYYY-MM-DD format")
-    parser.add_argument("--concat_with_latest_csv", action="store_true", help="Whether to also concatenate with the latest CSV from GitHub releases")
-    args = parser.parse_args()
-
-    compressed_dir = OUTPUT_DIR / "compressed"
-    date_dir = compressed_dir / args.date
-
-    parquet_files = sorted(date_dir.glob("*.parquet"))
-    df = None
-    if parquet_files: # TODO: This logic could be updated slightly.
-        print(f"No parquet files found in {date_dir}")
-
-        frames = [pl.read_parquet(p) for p in parquet_files]
-        df = pl.concat(frames, how="vertical", rechunk=True)
-
-        df = df.sort(["time", "icao"])
-        df = df.select(CORRECT_ORDER_OF_COLUMNS)
-        
-        output_path = OUTPUT_DIR / f"openairframes_adsb_{args.date}.parquet"
-        print(f"Writing combined parquet to {output_path} with {df.height} rows")
-        df.write_parquet(output_path)
-
-        csv_output_path = OUTPUT_DIR / f"openairframes_adsb_{args.date}.csv.gz"
-        print(f"Writing combined csv.gz to {csv_output_path} with {df.height} rows")
-        df.write_csv(csv_output_path, compression="gzip")
-
-    if args.concat_with_latest_csv:
-        print("Loading latest CSV from GitHub releases to concatenate with...")
-        from src.get_latest_release import get_latest_aircraft_adsb_csv_df
-        from datetime import datetime
-        
-        df_latest_csv, csv_start_date, csv_end_date = get_latest_aircraft_adsb_csv_df()
-        
-        # Compare dates: end_date is exclusive, so if csv_end_date > args.date, 
-        # the latest CSV already includes this day's data
-        csv_end_dt = datetime.strptime(csv_end_date, "%Y-%m-%d")
-        args_dt = datetime.strptime(args.date, "%Y-%m-%d")
-        
-        if df is None or csv_end_dt >= args_dt:
-            print(f"Latest CSV already includes data through {args.date} (end_date={csv_end_date} is exclusive)")
-            print("Writing latest CSV directly without concatenation to avoid duplicates")
-            os.makedirs(OUTPUT_DIR, exist_ok=True)
-            final_csv_output_path = OUTPUT_DIR / f"openairframes_adsb_{csv_start_date}_{csv_end_date}.csv.gz"
-            df_latest_csv = df_latest_csv.select(CORRECT_ORDER_OF_COLUMNS)
-            df_latest_csv.write_csv(final_csv_output_path, compression="gzip")
-        else:
-            print(f"Concatenating latest CSV (through {csv_end_date}) with new data ({args.date})")
-            # Ensure column order matches before concatenating
-            df_latest_csv = df_latest_csv.select(CORRECT_ORDER_OF_COLUMNS)
-            from src.adsb.compress_adsb_to_aircraft_data import concat_compressed_dfs
-            df_final = concat_compressed_dfs(df_latest_csv, df)
-            df_final = df_final.select(CORRECT_ORDER_OF_COLUMNS)
-            final_csv_output_path = OUTPUT_DIR / f"openairframes_adsb_{csv_start_date}_{args.date}.csv.gz"
-            df_final.write_csv(final_csv_output_path, compression="gzip")
-        print(f"Final CSV written to {final_csv_output_path}")
-
-if __name__ == "__main__":
-    main()

src/adsb/download_adsb_data_to_parquet.py

@@ -1,33 +1,42 @@
 """
 Downloads adsb.lol data and writes to Parquet files.
 
-This file contains utility functions for downloading and processing adsb.lol trace data.
-Used by the historical ADS-B processing pipeline.
+Usage:
+    python -m src.process_historical_adsb_data.download_to_parquet 2025-01-01 2025-01-02
+
+This will download trace data for the specified date range and output Parquet files.
+
+This file is self-contained and does not import from other project modules.
 """
-import datetime as dt
+import gc
+import glob
 import gzip
-import os
-import re
 import resource
 import shutil
-import signal
-import subprocess
 import sys
-import urllib.error
-import urllib.request
-from datetime import datetime
+import logging
 import time
+import re
+import signal
+import concurrent.futures
+import subprocess
+import os
+import argparse
+import datetime as dt
+from datetime import datetime, timedelta, timezone
+import urllib.request
+import urllib.error
+
 import orjson
 import pyarrow as pa
 import pyarrow.parquet as pq
-from pathlib import Path
 
 
 # ============================================================================
 # Configuration
 # ============================================================================
 
-OUTPUT_DIR = Path("./data/output")
+OUTPUT_DIR = "./data/output"
 os.makedirs(OUTPUT_DIR, exist_ok=True)
 
 PARQUET_DIR = os.path.join(OUTPUT_DIR, "parquet_output")
@@ -76,7 +85,7 @@ def _fetch_releases_from_repo(year: str, version_date: str) -> list:
     
     while True:
         max_retries = 10
-        retry_delay = 60*5
+        retry_delay = 60
         
         for attempt in range(1, max_retries + 1):
             try:
@@ -88,7 +97,7 @@ def _fetch_releases_from_repo(year: str, version_date: str) -> list:
                     else:
                         print(f"Failed to fetch releases (attempt {attempt}/{max_retries}): {response.status} {response.reason}")
                         if attempt < max_retries:
-                            print(f"Waiting {retry_delay} seconds before retry")
+                            print(f"Waiting {retry_delay} seconds before retry...")
                             time.sleep(retry_delay)
                         else:
                             print(f"Giving up after {max_retries} attempts")
@@ -96,7 +105,7 @@ def _fetch_releases_from_repo(year: str, version_date: str) -> list:
             except Exception as e:
                 print(f"Request exception (attempt {attempt}/{max_retries}): {e}")
                 if attempt < max_retries:
-                    print(f"Waiting {retry_delay} seconds before retry")
+                    print(f"Waiting {retry_delay} seconds before retry...")
                     time.sleep(retry_delay)
                 else:
                     print(f"Giving up after {max_retries} attempts")
@@ -123,105 +132,47 @@ def fetch_releases(version_date: str) -> list:
     # For last day of year, also check next year's repo if nothing found
     if not releases and version_date.endswith(".12.31"):
         next_year = str(int(year) + 1)
-        print(f"No releases found for {version_date} in {year} repo, checking {next_year} repo")
+        print(f"No releases found for {version_date} in {year} repo, checking {next_year} repo...")
         releases = _fetch_releases_from_repo(next_year, version_date)
     
     return releases
 
 
-def download_asset(asset_url: str, file_path: str, expected_size: int | None = None) -> bool:
-    """Download a single release asset with size verification.
-    
-    Args:
-        asset_url: URL to download from
-        file_path: Local path to save to
-        expected_size: Expected file size in bytes (for verification)
-    
-    Returns:
-        True if download succeeded and size matches (if provided), False otherwise
-    """
+def download_asset(asset_url: str, file_path: str) -> bool:
+    """Download a single release asset."""
     os.makedirs(os.path.dirname(file_path) or OUTPUT_DIR, exist_ok=True)
     
-    # Check if file exists and has correct size
     if os.path.exists(file_path):
-        if expected_size is not None:
-            actual_size = os.path.getsize(file_path)
-            if actual_size == expected_size:
-                print(f"[SKIP] {file_path} already downloaded and verified ({actual_size} bytes).")
+        print(f"[SKIP] {file_path} already downloaded.")
+        return True
+    
+    print(f"Downloading {asset_url}...")
+    try:
+        signal.signal(signal.SIGALRM, timeout_handler)
+        signal.alarm(40)  # 40-second timeout
+        
+        req = urllib.request.Request(asset_url, headers=HEADERS)
+        with urllib.request.urlopen(req) as response:
+            signal.alarm(0)
+            
+            if response.status == 200:
+                with open(file_path, "wb") as file:
+                    while True:
+                        chunk = response.read(8192)
+                        if not chunk:
+                            break
+                        file.write(chunk)
+                print(f"Saved {file_path}")
                 return True
             else:
-                print(f"[WARN] {file_path} exists but size mismatch (expected {expected_size}, got {actual_size}). Re-downloading.")
-                os.remove(file_path)
-        else:
-            print(f"[SKIP] {file_path} already downloaded.")
-            return True
-    
-    max_retries = 2
-    retry_delay = 30
-    timeout_seconds = 140
-    
-    for attempt in range(1, max_retries + 1):
-        print(f"Downloading {asset_url} (attempt {attempt}/{max_retries})")
-        try:
-            req = urllib.request.Request(asset_url, headers=HEADERS)
-            with urllib.request.urlopen(req, timeout=timeout_seconds) as response:
-                if response.status == 200:
-                    with open(file_path, "wb") as file:
-                        while True:
-                            chunk = response.read(8192)
-                            if not chunk:
-                                break
-                            file.write(chunk)
-                    
-                    # Verify file size if expected_size was provided
-                    if expected_size is not None:
-                        actual_size = os.path.getsize(file_path)
-                        if actual_size != expected_size:
-                            print(f"[ERROR] Size mismatch for {file_path}: expected {expected_size} bytes, got {actual_size} bytes")
-                            os.remove(file_path)
-                            if attempt < max_retries:
-                                print(f"Waiting {retry_delay} seconds before retry")
-                                time.sleep(retry_delay)
-                                continue
-                            return False
-                        print(f"Saved {file_path} ({actual_size} bytes, verified)")
-                    else:
-                        print(f"Saved {file_path}")
-                    return True
-                else:
-                    print(f"Failed to download {asset_url}: {response.status} {response.msg}")
-                    if attempt < max_retries:
-                        print(f"Waiting {retry_delay} seconds before retry")
-                        time.sleep(retry_delay)
-                    else:
-                        return False
-        except urllib.error.HTTPError as e:
-            if e.code == 404:
-                print(f"404 Not Found: {asset_url}")
-                raise Exception(f"Asset not found (404): {asset_url}")
-            else:
-                print(f"HTTP error occurred (attempt {attempt}/{max_retries}): {e.code} {e.reason}")
-                if attempt < max_retries:
-                    print(f"Waiting {retry_delay} seconds before retry")
-                    time.sleep(retry_delay)
-                else:
-                    return False
-        except urllib.error.URLError as e:
-            print(f"URL/Timeout error (attempt {attempt}/{max_retries}): {e}")
-            if attempt < max_retries:
-                print(f"Waiting {retry_delay} seconds before retry")
-                time.sleep(retry_delay)
-            else:
+                print(f"Failed to download {asset_url}: {response.status} {response.msg}")
                 return False
-        except Exception as e:
-            print(f"An error occurred (attempt {attempt}/{max_retries}): {e}")
-            if attempt < max_retries:
-                print(f"Waiting {retry_delay} seconds before retry")
-                time.sleep(retry_delay)
-            else:
-                return False
-    
-    return False
+    except DownloadTimeoutException as e:
+        print(f"Download aborted for {asset_url}: {e}")
+        return False
+    except Exception as e:
+        print(f"An error occurred while downloading {asset_url}: {e}")
+        return False
 
 
 def extract_split_archive(file_paths: list, extract_dir: str) -> bool:
@@ -260,6 +211,7 @@ def extract_split_archive(file_paths: list, extract_dir: str) -> bool:
             stdin=cat_proc.stdout,
             stdout=subprocess.PIPE,
             stderr=subprocess.PIPE,
+            check=True
         )
         cat_proc.stdout.close()
         cat_stderr = cat_proc.stderr.read().decode() if cat_proc.stderr else ""
@@ -268,24 +220,6 @@ def extract_split_archive(file_paths: list, extract_dir: str) -> bool:
         if cat_stderr:
             print(f"cat stderr: {cat_stderr}")
         
-        tar_stderr = result.stderr.decode() if result.stderr else ""
-        if result.returncode != 0:
-            # GNU tar exits non-zero for format issues that BSD tar silently
-            # tolerates (e.g. trailing junk after the last valid entry).
-            # Check whether files were actually extracted before giving up.
-            extracted_items = os.listdir(extract_dir)
-            if extracted_items:
-                print(f"[WARN] tar exited {result.returncode} but extracted "
-                      f"{len(extracted_items)} items — treating as success")
-                if tar_stderr:
-                    print(f"tar stderr: {tar_stderr}")
-            else:
-                print(f"Failed to extract split archive (tar exit {result.returncode})")
-                if tar_stderr:
-                    print(f"tar stderr: {tar_stderr}")
-                shutil.rmtree(extract_dir, ignore_errors=True)
-                return False
-        
         print(f"Successfully extracted archive to {extract_dir}")
         
         # Delete tar files immediately after extraction
@@ -302,9 +236,11 @@ def extract_split_archive(file_paths: list, extract_dir: str) -> bool:
         print(f"Disk space after tar deletion: {free_gb:.1f}GB free")
         
         return True
-    except Exception as e:
+    except subprocess.CalledProcessError as e:
+        stderr_output = e.stderr.decode() if e.stderr else ""
         print(f"Failed to extract split archive: {e}")
-        shutil.rmtree(extract_dir, ignore_errors=True)
+        if stderr_output:
+            print(f"tar stderr: {stderr_output}")
         return False
 
 
@@ -468,6 +404,8 @@ COLUMNS = [
 
 OS_CPU_COUNT = os.cpu_count() or 1
 MAX_WORKERS = OS_CPU_COUNT if OS_CPU_COUNT > 4 else 1
+CHUNK_SIZE = MAX_WORKERS * 500  # Reduced for lower RAM usage
+BATCH_SIZE = 250_000  # Fixed size for predictable memory usage (~500MB per batch)
 
 # PyArrow schema for efficient Parquet writing
 PARQUET_SCHEMA = pa.schema([
@@ -555,6 +493,217 @@ def collect_trace_files_with_find(root_dir):
     return trace_dict
 
 
+def generate_version_dates(start_date: str, end_date: str) -> list:
+    """Generate a list of dates from start_date to end_date inclusive."""
+    start = datetime.strptime(start_date, "%Y-%m-%d")
+    end = datetime.strptime(end_date, "%Y-%m-%d")
+    delta = end - start
+    return [start + timedelta(days=i) for i in range(delta.days + 1)]
+
+
+def safe_process(fp):
+    """Safely process a file, returning empty list on error."""
+    try:
+        return process_file(fp)
+    except Exception as e:
+        logging.error(f"Error processing {fp}: {e}")
+        return []
+
+
+def rows_to_arrow_table(rows: list) -> pa.Table:
+    """Convert list of rows to a PyArrow Table directly (no pandas)."""
+    # Transpose rows into columns
+    columns = list(zip(*rows))
+    
+    # Build arrays for each column according to schema
+    arrays = []
+    for i, field in enumerate(PARQUET_SCHEMA):
+        col_data = list(columns[i]) if i < len(columns) else [None] * len(rows)
+        arrays.append(pa.array(col_data, type=field.type))
+    
+    return pa.Table.from_arrays(arrays, schema=PARQUET_SCHEMA)
+
+
+def write_batch_to_parquet(rows: list, version_date: str, batch_idx: int):
+    """Write a batch of rows to a Parquet file."""
+    if not rows:
+        return
+    
+    table = rows_to_arrow_table(rows)
+    
+    parquet_path = os.path.join(PARQUET_DIR, f"{version_date}_batch_{batch_idx:04d}.parquet")
+    
+    pq.write_table(table, parquet_path, compression='snappy')
+    
+    print(f"Written parquet batch {batch_idx} ({len(rows)} rows) | {get_resource_usage()}")
+
+
+def merge_parquet_files(version_date: str, delete_batches: bool = True):
+    """Merge all batch parquet files for a version_date into a single file using streaming."""
+    pattern = os.path.join(PARQUET_DIR, f"{version_date}_batch_*.parquet")
+    batch_files = sorted(glob.glob(pattern))
+    
+    if not batch_files:
+        print(f"No batch files found for {version_date}")
+        return None
+    
+    print(f"Merging {len(batch_files)} batch files for {version_date} (streaming)...")
+    
+    merged_path = os.path.join(PARQUET_DIR, f"{version_date}.parquet")
+    total_rows = 0
+    
+    # Stream write: read one batch at a time to minimize RAM usage
+    writer = None
+    try:
+        for i, f in enumerate(batch_files):
+            table = pq.read_table(f)
+            total_rows += table.num_rows
+            
+            if writer is None:
+                writer = pq.ParquetWriter(merged_path, table.schema, compression='snappy')
+            
+            writer.write_table(table)
+            
+            # Delete batch file immediately after reading to free disk space
+            if delete_batches:
+                os.remove(f)
+            
+            # Free memory
+            del table
+            if (i + 1) % 10 == 0:
+                gc.collect()
+                print(f"  Merged {i + 1}/{len(batch_files)} batches... | {get_resource_usage()}")
+    finally:
+        if writer is not None:
+            writer.close()
+    
+    print(f"Merged parquet file written to {merged_path} ({total_rows} total rows) | {get_resource_usage()}")
+    
+    if delete_batches:
+        print(f"Deleted {len(batch_files)} batch files during merge")
+    
+    gc.collect()
+    return merged_path
+
+
+def process_version_date(version_date: str, keep_folders: bool = False):
+    """Download, extract, and process trace files for a single version date."""
+    print(f"\nProcessing version_date: {version_date}")
+    extract_dir = os.path.join(OUTPUT_DIR, f"{version_date}-planes-readsb-prod-0.tar_0")
+    
+    def collect_trace_files_for_version_date(vd):
+        releases = fetch_releases(vd)
+        if len(releases) == 0:
+            print(f"No releases found for {vd}.")
+            return None
+        
+        # Prefer non-tmp releases; only use tmp if no normal releases exist
+        normal_releases = [r for r in releases if "tmp" not in r["tag_name"]]
+        tmp_releases = [r for r in releases if "tmp" in r["tag_name"]]
+        releases = normal_releases if normal_releases else tmp_releases
+        print(f"Using {'normal' if normal_releases else 'tmp'} releases ({len(releases)} found)")
+        
+        downloaded_files = []
+        for release in releases:
+            tag_name = release["tag_name"]
+            print(f"Processing release: {tag_name}")
+
+            # Only download prod-0 if available, else prod-0tmp
+            assets = release.get("assets", [])
+            normal_assets = [
+                a for a in assets
+                if "planes-readsb-prod-0." in a["name"] and "tmp" not in a["name"]
+            ]
+            tmp_assets = [
+                a for a in assets
+                if "planes-readsb-prod-0tmp" in a["name"]
+            ]
+            use_assets = normal_assets if normal_assets else tmp_assets
+
+            for asset in use_assets:
+                asset_name = asset["name"]
+                asset_url = asset["browser_download_url"]
+                file_path = os.path.join(OUTPUT_DIR, asset_name)
+                result = download_asset(asset_url, file_path)
+                if result:
+                    downloaded_files.append(file_path)
+
+        extract_split_archive(downloaded_files, extract_dir)
+        return collect_trace_files_with_find(extract_dir)
+
+    # Check if files already exist
+    pattern = os.path.join(OUTPUT_DIR, f"{version_date}-planes-readsb-prod-0*")
+    matches = [p for p in glob.glob(pattern) if os.path.isfile(p)]
+    
+    if matches:
+        print(f"Found existing files for {version_date}:")
+        # Prefer non-tmp slices when reusing existing files
+        normal_matches = [
+            p for p in matches
+            if "-planes-readsb-prod-0." in os.path.basename(p)
+            and "tmp" not in os.path.basename(p)
+        ]
+        downloaded_files = normal_matches if normal_matches else matches
+        
+        extract_split_archive(downloaded_files, extract_dir)
+        trace_files = collect_trace_files_with_find(extract_dir)
+    else:
+        trace_files = collect_trace_files_for_version_date(version_date)
+    
+    if trace_files is None or len(trace_files) == 0:
+        print(f"No trace files found for version_date: {version_date}")
+        return 0
+    
+    file_list = list(trace_files.values())
+    
+    start_time = time.perf_counter()
+    total_num_rows = 0
+    batch_rows = []
+    batch_idx = 0
+    
+    # Process files in chunks
+    for offset in range(0, len(file_list), CHUNK_SIZE):
+        chunk = file_list[offset:offset + CHUNK_SIZE]
+        with concurrent.futures.ProcessPoolExecutor(max_workers=MAX_WORKERS) as process_executor:
+            for rows in process_executor.map(safe_process, chunk):
+                if not rows:
+                    continue
+                batch_rows.extend(rows)
+                
+                if len(batch_rows) >= BATCH_SIZE:
+                    total_num_rows += len(batch_rows)
+                    write_batch_to_parquet(batch_rows, version_date, batch_idx)
+                    batch_idx += 1
+                    batch_rows = []
+                    
+                    elapsed = time.perf_counter() - start_time
+                    speed = total_num_rows / elapsed if elapsed > 0 else 0
+                    print(f"[{version_date}] processed {total_num_rows} rows in {elapsed:.2f}s ({speed:.2f} rows/s)")
+        
+        gc.collect()
+    
+    # Final batch
+    if batch_rows:
+        total_num_rows += len(batch_rows)
+        write_batch_to_parquet(batch_rows, version_date, batch_idx)
+        elapsed = time.perf_counter() - start_time
+        speed = total_num_rows / elapsed if elapsed > 0 else 0
+        print(f"[{version_date}] processed {total_num_rows} rows in {elapsed:.2f}s ({speed:.2f} rows/s)")
+    
+    print(f"Total rows processed for version_date {version_date}: {total_num_rows}")
+    
+    # Clean up extracted directory immediately after processing (before merging parquet files)
+    if not keep_folders and os.path.isdir(extract_dir):
+        print(f"Deleting extraction directory with 100,000+ files: {extract_dir}")
+        shutil.rmtree(extract_dir)
+        print(f"Successfully deleted extraction directory: {extract_dir} | {get_resource_usage()}")
+    
+    # Merge batch files into a single parquet file
+    merge_parquet_files(version_date, delete_batches=True)
+    
+    return total_num_rows
+
+
 def create_parquet_for_day(day, keep_folders: bool = False):
     """Create parquet file for a single day.
     
@@ -578,10 +727,42 @@ def create_parquet_for_day(day, keep_folders: bool = False):
         print(f"Parquet file already exists: {parquet_path}")
         return parquet_path
     
-    print(f"Creating parquet for {version_date}")
+    print(f"Creating parquet for {version_date}...")
     rows_processed = process_version_date(version_date, keep_folders)
     
     if rows_processed > 0 and parquet_path.exists():
         return parquet_path
     else:
         return None
+
+
+def main(start_date: str, end_date: str, keep_folders: bool = False):
+    """Main function to download and convert adsb.lol data to Parquet."""
+    version_dates = [f"v{date.strftime('%Y.%m.%d')}" for date in generate_version_dates(start_date, end_date)]
+    print(f"Processing dates: {version_dates}")
+    
+    total_rows_all = 0
+    for version_date in version_dates:
+        rows_processed = process_version_date(version_date, keep_folders)
+        total_rows_all += rows_processed
+    
+    print(f"\n=== Summary ===")
+    print(f"Total dates processed: {len(version_dates)}")
+    print(f"Total rows written to Parquet: {total_rows_all}")
+    print(f"Parquet files location: {PARQUET_DIR}")
+
+
+if __name__ == "__main__":
+    logging.basicConfig(level=logging.INFO, stream=sys.stdout, force=True)
+    
+    parser = argparse.ArgumentParser(
+        description="Download adsb.lol data and write to Parquet files"
+    )
+    parser.add_argument("start_date", type=str, help="Start date in YYYY-MM-DD format")
+    parser.add_argument("end_date", type=str, help="End date in YYYY-MM-DD format")
+    parser.add_argument("--keep-folders", action="store_true", 
+                        help="Keep extracted folders after processing")
+    
+    args = parser.parse_args()
+    
+    main(args.start_date, args.end_date, args.keep_folders)

src/adsb/download_and_list_icaos.py

@@ -1,7 +1,9 @@
 """
-Downloads and extracts adsb.lol tar files for a single day, then lists all ICAO folders.
+Downloads and extracts adsb.lol tar files, then lists all ICAO folders.
 This is the first step of the map-reduce pipeline.
 
+Supports both single-day (daily) and multi-day (historical) modes.
+
 Outputs:
 - Extracted trace files in data/output/{version_date}-planes-readsb-prod-0.tar_0/
 - ICAO manifest at data/output/icao_manifest_{date}.txt
@@ -23,6 +25,11 @@ from src.adsb.download_adsb_data_to_parquet import (
 )
 
 
+def get_target_day() -> datetime:
+    """Get yesterday's date (the day we're processing)."""
+    return datetime.utcnow() - timedelta(days=1)
+
+
 def download_and_extract(version_date: str) -> str | None:
     """Download and extract tar files, return extract directory path."""
     extract_dir = os.path.join(OUTPUT_DIR, f"{version_date}-planes-readsb-prod-0.tar_0")
@@ -77,9 +84,8 @@ def download_and_extract(version_date: str) -> str | None:
             for asset in use_assets:
                 asset_name = asset["name"]
                 asset_url = asset["browser_download_url"]
-                asset_size = asset.get("size")  # Get expected file size
                 file_path = os.path.join(OUTPUT_DIR, asset_name)
-                if download_asset(asset_url, file_path, expected_size=asset_size):
+                if download_asset(asset_url, file_path):
                     downloaded_files.append(file_path)
     
     if not downloaded_files:
@@ -100,6 +106,21 @@ def list_icao_folders(extract_dir: str) -> list[str]:
     return icaos
 
 
+def write_manifest(icaos: list[str], manifest_id: str) -> str:
+    """Write ICAO list to manifest file.
+    
+    Args:
+        icaos: List of ICAO codes
+        manifest_id: Identifier for manifest file (date or date range)
+    """
+    manifest_path = os.path.join(OUTPUT_DIR, f"icao_manifest_{manifest_id}.txt")
+    with open(manifest_path, "w") as f:
+        for icao in sorted(icaos):
+            f.write(f"{icao}\n")
+    print(f"Wrote manifest with {len(icaos)} ICAOs to {manifest_path}")
+    return manifest_path
+
+
 def process_single_day(target_day: datetime) -> tuple[str | None, list[str]]:
     """Process a single day: download, extract, list ICAOs.
     
@@ -114,50 +135,82 @@ def process_single_day(target_day: datetime) -> tuple[str | None, list[str]]:
     extract_dir = download_and_extract(version_date)
     if not extract_dir:
         print(f"Failed to download/extract data for {date_str}")
-        raise Exception(f"No data available for {date_str}")
+        return None, []
     
     icaos = list_icao_folders(extract_dir)
     print(f"Found {len(icaos)} ICAOs for {date_str}")
     
     return extract_dir, icaos
 
-from pathlib import Path
-import tarfile
-NUMBER_PARTS = 4
-def split_folders_into_gzip_archives(extract_dir: Path, tar_output_dir: Path, icaos: list[str], parts = NUMBER_PARTS) -> list[str]:
-    traces_dir = extract_dir / "traces"
-    buckets = sorted(traces_dir.iterdir())
-    tars = []
-    for i in range(parts):
-        tar_path = tar_output_dir / f"{tar_output_dir.name}_part_{i}.tar.gz"
-        tars.append(tarfile.open(tar_path, "w:gz"))
-    for idx, bucket_path in enumerate(buckets):
-        tar_idx = idx % parts
-        tars[tar_idx].add(bucket_path, arcname=bucket_path.name)
-    for tar in tars:
-        tar.close()
+
+def process_date_range(start_date: datetime, end_date: datetime) -> set[str]:
+    """Process multiple days: download, extract, combine ICAO lists.
+    
+    Args:
+        start_date: Start date (inclusive)
+        end_date: End date (inclusive)
+    
+    Returns:
+        Combined set of all ICAOs across the date range
+    """
+    all_icaos: set[str] = set()
+    current = start_date
+    
+    # Both start and end are inclusive
+    while current <= end_date:
+        _, icaos = process_single_day(current)
+        all_icaos.update(icaos)
+        current += timedelta(days=1)
+    
+    return all_icaos
 
 
 def main():
-    parser = argparse.ArgumentParser(description="Download and list ICAOs from adsb.lol data for a single day")
+    parser = argparse.ArgumentParser(description="Download and list ICAOs from adsb.lol data")
     parser.add_argument("--date", type=str, help="Single date in YYYY-MM-DD format (default: yesterday)")
+    parser.add_argument("--start-date", type=str, help="Start date for range (YYYY-MM-DD)")
+    parser.add_argument("--end-date", type=str, help="End date for range (YYYY-MM-DD)")
     args = parser.parse_args()
     
-    target_day = datetime.strptime(args.date, "%Y-%m-%d")
-    date_str = target_day.strftime("%Y-%m-%d")
-    tar_output_dir = Path(f"./data/output/adsb_archives/{date_str}")
-    
-    extract_dir, icaos = process_single_day(target_day)
-    extract_dir = Path(extract_dir)
-    print(extract_dir)
-    tar_output_dir.mkdir(parents=True, exist_ok=True)
-    split_folders_into_gzip_archives(extract_dir, tar_output_dir, icaos)
-    if not icaos:
-        print("No ICAOs found")
-        sys.exit(1)
-    
-    print(f"\nDone! Extract dir: {extract_dir}")
-    print(f"Total ICAOs: {len(icaos)}")
+    # Determine mode: single day or date range
+    if args.start_date and args.end_date:
+        # Historical mode: process date range
+        start_date = datetime.strptime(args.start_date, "%Y-%m-%d")
+        end_date = datetime.strptime(args.end_date, "%Y-%m-%d")
+        
+        print(f"Processing date range: {args.start_date} to {args.end_date}")
+        
+        all_icaos = process_date_range(start_date, end_date)
+        
+        if not all_icaos:
+            print("No ICAOs found in date range")
+            sys.exit(1)
+        
+        # Write combined manifest with range identifier
+        manifest_id = f"{args.start_date}_{args.end_date}"
+        write_manifest(list(all_icaos), manifest_id)
+        
+        print(f"\nDone! Total ICAOs: {len(all_icaos)}")
+        
+    else:
+        # Daily mode: single day
+        if args.date:
+            target_day = datetime.strptime(args.date, "%Y-%m-%d")
+        else:
+            target_day = get_target_day()
+        
+        date_str = target_day.strftime("%Y-%m-%d")
+        
+        extract_dir, icaos = process_single_day(target_day)
+        
+        if not icaos:
+            print("No ICAOs found")
+            sys.exit(1)
+        
+        write_manifest(icaos, date_str)
+        
+        print(f"\nDone! Extract dir: {extract_dir}")
+        print(f"Total ICAOs: {len(icaos)}")
 
 
 if __name__ == "__main__":

src/adsb/historical_generate_matrix.py

@@ -41,7 +41,7 @@ def main() -> None:
     """Main entry point for GitHub Actions."""
     start_date = os.environ.get("INPUT_START_DATE")
     end_date = os.environ.get("INPUT_END_DATE")
-    chunk_days = int(os.environ.get("INPUT_CHUNK_DAYS", "1"))
+    chunk_days = int(os.environ.get("INPUT_CHUNK_DAYS", "7"))
     
     if not start_date or not end_date:
         print("ERROR: INPUT_START_DATE and INPUT_END_DATE must be set", file=sys.stderr)

src/adsb/main.py

@@ -1,78 +0,0 @@
-"""
-Main pipeline for processing ADS-B data from adsb.lol.
-
-Usage:
-    python -m src.adsb.main --date 2026-01-01
-    python -m src.adsb.main --start_date 2026-01-01 --end_date 2026-01-03
-"""
-import argparse
-import subprocess
-import sys
-from datetime import datetime, timedelta
-
-import polars as pl
-
-from src.adsb.download_and_list_icaos import NUMBER_PARTS
-
-
-def main():
-    parser = argparse.ArgumentParser(description="Process ADS-B data for a single day or date range")
-    parser.add_argument("--date", type=str, help="Single date in YYYY-MM-DD format")
-    parser.add_argument("--start_date", type=str, help="Start date (inclusive, YYYY-MM-DD)")
-    parser.add_argument("--end_date", type=str, help="End date (exclusive, YYYY-MM-DD)")
-    parser.add_argument("--concat_with_latest_csv", action="store_true", help="Also concatenate with latest CSV from GitHub releases")
-    args = parser.parse_args()
-
-    if args.date and (args.start_date or args.end_date):
-        raise SystemExit("Use --date or --start_date/--end_date, not both.")
-
-    if args.date:
-        start_date = datetime.strptime(args.date, "%Y-%m-%d")
-        end_date = start_date + timedelta(days=1)
-    else:
-        if not args.start_date or not args.end_date:
-            raise SystemExit("Provide --start_date and --end_date, or use --date.")
-        start_date = datetime.strptime(args.start_date, "%Y-%m-%d")
-        end_date = datetime.strptime(args.end_date, "%Y-%m-%d")
-
-    current = start_date
-    while current < end_date:
-        date_str = current.strftime("%Y-%m-%d")
-        print(f"Processing day: {date_str}")
-
-        # Download and split
-        subprocess.run([sys.executable, "-m", "src.adsb.download_and_list_icaos", "--date", date_str], check=True)
-
-        # Process parts
-        for part_id in range(NUMBER_PARTS):
-            subprocess.run([sys.executable, "-m", "src.adsb.process_icao_chunk", "--part-id", str(part_id), "--date", date_str], check=True)
-
-        # Concatenate
-        concat_cmd = [sys.executable, "-m", "src.adsb.concat_parquet_to_final", "--date", date_str]
-        if args.concat_with_latest_csv:
-            concat_cmd.append("--concat_with_latest_csv")
-        subprocess.run(concat_cmd, check=True)
-
-        current += timedelta(days=1)
-
-    if end_date - start_date > timedelta(days=1):
-        dates = []
-        cur = start_date
-        while cur < end_date:
-            dates.append(cur.strftime("%Y-%m-%d"))
-            cur += timedelta(days=1)
-        csv_files = [
-            f"data/outputs/openairframes_adsb_{d}_{d}.csv"
-            for d in dates
-        ]
-        frames = [pl.read_csv(p) for p in csv_files]
-        df = pl.concat(frames, how="vertical", rechunk=True)
-        output_path = f"data/outputs/openairframes_adsb_{start_date.strftime('%Y-%m-%d')}_{end_date.strftime('%Y-%m-%d')}.csv"
-        df.write_csv(output_path)
-        print(f"Wrote combined CSV: {output_path}")
-
-    print("Done")
-
-
-if __name__ == "__main__":
-    main()

src/adsb/process_icao_chunk.py

@@ -1,9 +1,18 @@
 """
-Processes trace files from a single archive part for a single day.
+Processes a chunk of ICAOs from pre-extracted trace files.
 This is the map phase of the map-reduce pipeline.
 
+Supports both single-day (daily) and multi-day (historical) modes.
+
+Expects extract_dir to already exist with trace files.
+Reads ICAO manifest to determine which ICAOs to process based on chunk-id.
+
 Usage:
-    python -m src.adsb.process_icao_chunk --part-id 1 --date 2026-01-01
+    # Daily mode (single day)
+    python -m src.adsb.process_icao_chunk --chunk-id 0 --total-chunks 4
+    
+    # Historical mode (date range)
+    python -m src.adsb.process_icao_chunk --chunk-id 0 --total-chunks 4 --start-date 2024-01-01 --end-date 2024-01-07
 """
 import gc
 import os
@@ -12,9 +21,6 @@ import argparse
 import time
 import concurrent.futures
 from datetime import datetime, timedelta
-import tarfile
-import tempfile
-import shutil
 
 import pyarrow as pa
 import pyarrow.parquet as pq
@@ -31,21 +37,72 @@ from src.adsb.download_adsb_data_to_parquet import (
 )
 
 
+CHUNK_OUTPUT_DIR = os.path.join(OUTPUT_DIR, "adsb_chunks")
+os.makedirs(CHUNK_OUTPUT_DIR, exist_ok=True)
+
 # Smaller batch size for memory efficiency
 BATCH_SIZE = 100_000
 
-def build_trace_file_map(archive_path: str) -> dict[str, str]:
-    """Build a map of ICAO -> trace file path by extracting tar.gz archive."""
-    print(f"Extracting {archive_path}...")
+
+def get_target_day() -> datetime:
+    """Get yesterday's date (the day we're processing)."""
+    return datetime.utcnow() - timedelta(days=1)
+
+
+def read_manifest(manifest_id: str) -> list[str]:
+    """Read ICAO manifest file.
     
-    temp_dir = tempfile.mkdtemp(prefix="adsb_extract_")
+    Args:
+        manifest_id: Either a date string (YYYY-MM-DD) or range string (YYYY-MM-DD_YYYY-MM-DD)
+    """
+    manifest_path = os.path.join(OUTPUT_DIR, f"icao_manifest_{manifest_id}.txt")
+    if not os.path.exists(manifest_path):
+        raise FileNotFoundError(f"Manifest not found: {manifest_path}")
     
-    with tarfile.open(archive_path, 'r:gz') as tar:
-        tar.extractall(path=temp_dir, filter='data')
+    with open(manifest_path, "r") as f:
+        icaos = [line.strip() for line in f if line.strip()]
+    return icaos
+
+
+def deterministic_hash(s: str) -> int:
+    """Return a deterministic hash for a string (unlike Python's hash() which is randomized)."""
+    # Use sum of byte values - simple but deterministic
+    return sum(ord(c) for c in s)
+
+
+def get_chunk_icaos(icaos: list[str], chunk_id: int, total_chunks: int) -> list[str]:
+    """Get the subset of ICAOs for this chunk based on deterministic hash partitioning."""
+    return [icao for icao in icaos if deterministic_hash(icao) % total_chunks == chunk_id]
+
+
+def build_trace_file_map(extract_dir: str) -> dict[str, str]:
+    """Build a map of ICAO -> trace file path using find command."""
+    print(f"Building trace file map from {extract_dir}...")
     
-    trace_map = collect_trace_files_with_find(temp_dir)
+    # Debug: check what's in extract_dir
+    if os.path.isdir(extract_dir):
+        items = os.listdir(extract_dir)[:10]
+        print(f"First 10 items in extract_dir: {items}")
+        # Check if there are subdirectories
+        for item in items[:3]:
+            subpath = os.path.join(extract_dir, item)
+            if os.path.isdir(subpath):
+                subitems = os.listdir(subpath)[:5]
+                print(f"  Contents of {item}/: {subitems}")
+    
+    trace_map = collect_trace_files_with_find(extract_dir)
     print(f"Found {len(trace_map)} trace files")
     
+    if len(trace_map) == 0:
+        # Debug: try manual find
+        import subprocess
+        result = subprocess.run(
+            ['find', extract_dir, '-type', 'f', '-name', 'trace_full_*'],
+            capture_output=True, text=True
+        )
+        print(f"Manual find output (first 500 chars): {result.stdout[:500]}")
+        print(f"Manual find stderr: {result.stderr[:200]}")
+    
     return trace_map
 
 
@@ -68,13 +125,42 @@ def rows_to_table(rows: list) -> pa.Table:
 
 
 def process_chunk(
-    trace_files: list[str],
-    part_id: int,
-    date_str: str,
+    chunk_id: int,
+    total_chunks: int,
+    trace_map: dict[str, str],
+    icaos: list[str],
+    output_id: str,
 ) -> str | None:
-    """Process trace files and write to a single parquet file."""
+    """Process a chunk of ICAOs and write to parquet.
     
-    output_path = os.path.join(PARQUET_DIR, f"part_{part_id}_{date_str}.parquet")
+    Args:
+        chunk_id: This chunk's ID (0-indexed)
+        total_chunks: Total number of chunks
+        trace_map: Map of ICAO -> trace file path
+        icaos: Full list of ICAOs from manifest
+        output_id: Identifier for output file (date or date range)
+    """
+    chunk_icaos = get_chunk_icaos(icaos, chunk_id, total_chunks)
+    print(f"Chunk {chunk_id}/{total_chunks}: Processing {len(chunk_icaos)} ICAOs")
+    
+    if not chunk_icaos:
+        print(f"Chunk {chunk_id}: No ICAOs to process")
+        return None
+    
+    # Get trace file paths from the map
+    trace_files = []
+    for icao in chunk_icaos:
+        if icao in trace_map:
+            trace_files.append(trace_map[icao])
+    
+    print(f"Chunk {chunk_id}: Found {len(trace_files)} trace files")
+    
+    if not trace_files:
+        print(f"Chunk {chunk_id}: No trace files found")
+        return None
+    
+    # Process files and write parquet in batches
+    output_path = os.path.join(CHUNK_OUTPUT_DIR, f"chunk_{chunk_id}_{output_id}.parquet")
     
     start_time = time.perf_counter()
     total_rows = 0
@@ -82,8 +168,7 @@ def process_chunk(
     writer = None
     
     try:
-        writer = pq.ParquetWriter(output_path, PARQUET_SCHEMA, compression='snappy')
-        
+        # Process in parallel batches
         files_per_batch = MAX_WORKERS * 100
         for offset in range(0, len(trace_files), files_per_batch):
             batch_files = trace_files[offset:offset + files_per_batch]
@@ -93,72 +178,166 @@ def process_chunk(
                     if rows:
                         batch_rows.extend(rows)
                         
+                        # Write when batch is full
                         if len(batch_rows) >= BATCH_SIZE:
-                            writer.write_table(rows_to_table(batch_rows))
+                            table = rows_to_table(batch_rows)
                             total_rows += len(batch_rows)
+                            
+                            if writer is None:
+                                writer = pq.ParquetWriter(output_path, PARQUET_SCHEMA, compression='snappy')
+                            writer.write_table(table)
+                            
                             batch_rows = []
+                            del table
                             gc.collect()
+                            
+                            elapsed = time.perf_counter() - start_time
+                            print(f"Chunk {chunk_id}: {total_rows} rows, {elapsed:.1f}s | {get_resource_usage()}")
+            
             gc.collect()
         
+        # Write remaining rows
         if batch_rows:
-            writer.write_table(rows_to_table(batch_rows))
+            table = rows_to_table(batch_rows)
             total_rows += len(batch_rows)
+            
+            if writer is None:
+                writer = pq.ParquetWriter(output_path, PARQUET_SCHEMA, compression='snappy')
+            writer.write_table(table)
+            del table
     
     finally:
         if writer:
             writer.close()
     
-    print(f"Part {part_id}: Done! {total_rows} rows in {time.perf_counter() - start_time:.1f}s | {get_resource_usage()}")
+    elapsed = time.perf_counter() - start_time
+    print(f"Chunk {chunk_id}: Done! {total_rows} rows in {elapsed:.1f}s | {get_resource_usage()}")
     
-    return output_path if total_rows > 0 else None
+    if total_rows > 0:
+        return output_path
+    return None
+
+
+def process_single_day(
+    chunk_id: int,
+    total_chunks: int,
+    target_day: datetime,
+) -> str | None:
+    """Process a single day for this chunk."""
+    date_str = target_day.strftime("%Y-%m-%d")
+    version_date = f"v{target_day.strftime('%Y.%m.%d')}"
+    
+    extract_dir = os.path.join(OUTPUT_DIR, f"{version_date}-planes-readsb-prod-0.tar_0")
+    
+    if not os.path.isdir(extract_dir):
+        print(f"Extract directory not found: {extract_dir}")
+        return None
+    
+    trace_map = build_trace_file_map(extract_dir)
+    if not trace_map:
+        print("No trace files found")
+        return None
+    
+    icaos = read_manifest(date_str)
+    print(f"Total ICAOs in manifest: {len(icaos)}")
+    
+    return process_chunk(chunk_id, total_chunks, trace_map, icaos, date_str)
+
+
+def process_date_range(
+    chunk_id: int,
+    total_chunks: int,
+    start_date: datetime,
+    end_date: datetime,
+) -> str | None:
+    """Process a date range for this chunk.
+    
+    Combines trace files from all days in the range.
+    
+    Args:
+        chunk_id: This chunk's ID (0-indexed)
+        total_chunks: Total number of chunks
+        start_date: Start date (inclusive)
+        end_date: End date (inclusive)
+    """
+    start_str = start_date.strftime("%Y-%m-%d")
+    end_str = end_date.strftime("%Y-%m-%d")
+    manifest_id = f"{start_str}_{end_str}"
+    
+    print(f"Processing date range: {start_str} to {end_str}")
+    
+    # Build combined trace map from all days
+    combined_trace_map: dict[str, str] = {}
+    current = start_date
+    
+    # Both start and end are inclusive
+    while current <= end_date:
+        version_date = f"v{current.strftime('%Y.%m.%d')}"
+        extract_dir = os.path.join(OUTPUT_DIR, f"{version_date}-planes-readsb-prod-0.tar_0")
+        
+        if os.path.isdir(extract_dir):
+            trace_map = build_trace_file_map(extract_dir)
+            # Later days override earlier days (use most recent trace file)
+            combined_trace_map.update(trace_map)
+            print(f"  {current.strftime('%Y-%m-%d')}: {len(trace_map)} trace files")
+        else:
+            print(f"  {current.strftime('%Y-%m-%d')}: no extract directory")
+        
+        current += timedelta(days=1)
+    
+    if not combined_trace_map:
+        print("No trace files found in date range")
+        return None
+    
+    print(f"Combined trace map: {len(combined_trace_map)} ICAOs")
+    
+    icaos = read_manifest(manifest_id)
+    print(f"Total ICAOs in manifest: {len(icaos)}")
+    
+    return process_chunk(chunk_id, total_chunks, combined_trace_map, icaos, manifest_id)
 
-from pathlib import Path
 
 def main():
-    parser = argparse.ArgumentParser(description="Process a single archive part for a day")
-    parser.add_argument("--part-id", type=int, required=True, help="Part ID (1-indexed)")
-    parser.add_argument("--date", type=str, required=True, help="Date in YYYY-MM-DD format")
+    parser = argparse.ArgumentParser(description="Process a chunk of ICAOs")
+    parser.add_argument("--chunk-id", type=int, required=True, help="Chunk ID (0-indexed)")
+    parser.add_argument("--total-chunks", type=int, required=True, help="Total number of chunks")
+    parser.add_argument("--date", type=str, help="Single date in YYYY-MM-DD format (default: yesterday)")
+    parser.add_argument("--start-date", type=str, help="Start date for range (YYYY-MM-DD)")
+    parser.add_argument("--end-date", type=str, help="End date for range (YYYY-MM-DD)")
     args = parser.parse_args()
     
-    print(f"Processing part {args.part_id} for {args.date}")
+    print(f"Processing chunk {args.chunk_id}/{args.total_chunks}")
+    print(f"OUTPUT_DIR: {OUTPUT_DIR}")
+    print(f"CHUNK_OUTPUT_DIR: {CHUNK_OUTPUT_DIR}")
+    print(f"Resource usage at start: {get_resource_usage()}")
     
-    # Get specific archive file for this part
-    archive_dir = os.path.join(OUTPUT_DIR, "adsb_archives", args.date)
-    archive_path = os.path.join(archive_dir, f"{args.date}_part_{args.part_id}.tar.gz")
+    # Debug: List what's in OUTPUT_DIR
+    print(f"\nContents of {OUTPUT_DIR}:")
+    if os.path.isdir(OUTPUT_DIR):
+        for item in os.listdir(OUTPUT_DIR)[:20]:
+            print(f"  - {item}")
+    else:
+        print(f"  Directory does not exist!")
     
-    if not os.path.isfile(archive_path):
-        print(f"ERROR: Archive not found: {archive_path}")
-        if os.path.isdir(archive_dir):
-            print(f"Files in {archive_dir}: {os.listdir(archive_dir)}")
+    # Determine mode: single day or date range
+    if args.start_date and args.end_date:
+        # Historical mode
+        start_date = datetime.strptime(args.start_date, "%Y-%m-%d")
+        end_date = datetime.strptime(args.end_date, "%Y-%m-%d")
+        output_path = process_date_range(args.chunk_id, args.total_chunks, start_date, end_date)
+    else:
+        # Daily mode
+        if args.date:
+            target_day = datetime.strptime(args.date, "%Y-%m-%d")
         else:
-            print(f"Directory does not exist: {archive_dir}")
-        sys.exit(1)
+            target_day = get_target_day()
+        output_path = process_single_day(args.chunk_id, args.total_chunks, target_day)
     
-    # Extract and collect trace files
-    trace_map = build_trace_file_map(archive_path)
-    all_trace_files = list(trace_map.values())
-    
-    print(f"Total trace files: {len(all_trace_files)}")
-    
-    # Process and write output
-    output_path = process_chunk(all_trace_files, args.part_id, args.date)
-    
-    from src.adsb.compress_adsb_to_aircraft_data import compress_parquet_part
-    df_compressed = compress_parquet_part(args.part_id, args.date)
-    
-    # Write parquet
-    df_compressed_output = OUTPUT_DIR / "compressed" / args.date/ f"part_{args.part_id}_{args.date}.parquet"
-    os.makedirs(df_compressed_output.parent, exist_ok=True)
-    df_compressed.write_parquet(df_compressed_output, compression='snappy')
-    
-    # Write CSV
-    csv_output = OUTPUT_DIR / "compressed" / args.date / f"part_{args.part_id}_{args.date}.csv"
-    df_compressed.write_csv(csv_output)
-    
-    print(f"Raw output: {output_path}" if output_path else "No raw output generated")
-    print(f"Compressed parquet: {df_compressed_output}")
-    print(f"Compressed CSV: {csv_output}")
+    if output_path:
+        print(f"Output: {output_path}")
+    else:
+        print("No output generated")
 
 
 if __name__ == "__main__":
-    main()
+    main()

src/adsb/requirements.reducer.txt

@@ -0,0 +1,2 @@
+polars>=1.0
+boto3>=1.34

src/adsb/requirements.worker.txt

@@ -0,0 +1,5 @@
+polars>=1.0
+pyarrow>=14.0
+orjson>=3.9
+boto3>=1.34
+zstandard>=0.22

src/adsb/run_local.py

@@ -0,0 +1,155 @@
+#!/usr/bin/env python3
+"""
+Run the full ADS-B processing pipeline locally.
+
+Downloads adsb.lol data, processes trace files, and outputs openairframes_adsb CSV.
+
+Usage:
+    # Single day (yesterday by default)
+    python -m src.adsb.run_local
+    
+    # Single day (specific date)
+    python -m src.adsb.run_local 2024-01-15
+    
+    # Date range (inclusive)
+    python -m src.adsb.run_local 2024-01-01 2024-01-07
+"""
+import argparse
+import os
+import subprocess
+import sys
+from datetime import datetime, timedelta
+
+
+def run_cmd(cmd: list[str], description: str) -> None:
+    """Run a command and exit on failure."""
+    print(f"\n>>> {' '.join(cmd)}")
+    result = subprocess.run(cmd)
+    if result.returncode != 0:
+        print(f"ERROR: {description} failed with exit code {result.returncode}")
+        sys.exit(result.returncode)
+
+
+def main():
+    parser = argparse.ArgumentParser(
+        description="Run full ADS-B processing pipeline locally",
+        usage="python -m src.adsb.run_local [start_date] [end_date]"
+    )
+    parser.add_argument(
+        "start_date",
+        nargs="?",
+        help="Start date (YYYY-MM-DD). Default: yesterday"
+    )
+    parser.add_argument(
+        "end_date",
+        nargs="?",
+        help="End date (YYYY-MM-DD, inclusive). If omitted, processes single day"
+    )
+    parser.add_argument(
+        "--chunks",
+        type=int,
+        default=4,
+        help="Number of parallel chunks (default: 4)"
+    )
+    parser.add_argument(
+        "--skip-base",
+        action="store_true",
+        help="Skip downloading and merging with base release"
+    )
+    args = parser.parse_args()
+
+    # Determine dates
+    if args.start_date:
+        start_date = datetime.strptime(args.start_date, "%Y-%m-%d")
+    else:
+        start_date = datetime.utcnow() - timedelta(days=1)
+    
+    end_date = None
+    if args.end_date:
+        end_date = datetime.strptime(args.end_date, "%Y-%m-%d")
+    
+    start_str = start_date.strftime("%Y-%m-%d")
+    end_str = end_date.strftime("%Y-%m-%d") if end_date else None
+    
+    print("=" * 60)
+    print("ADS-B Processing Pipeline")
+    print("=" * 60)
+    if end_str:
+        print(f"Date range: {start_str} to {end_str}")
+    else:
+        print(f"Date: {start_str}")
+    print(f"Chunks: {args.chunks}")
+    print("=" * 60)
+    
+    # Step 1: Download and extract
+    print("\n" + "=" * 60)
+    print("Step 1: Download and Extract")
+    print("=" * 60)
+    
+    if end_str:
+        cmd = ["python", "-m", "src.adsb.download_and_list_icaos",
+               "--start-date", start_str, "--end-date", end_str]
+    else:
+        cmd = ["python", "-m", "src.adsb.download_and_list_icaos",
+               "--date", start_str]
+    run_cmd(cmd, "Download and extract")
+    
+    # Step 2: Process chunks
+    print("\n" + "=" * 60)
+    print("Step 2: Process Chunks")
+    print("=" * 60)
+    
+    for chunk_id in range(args.chunks):
+        print(f"\n--- Chunk {chunk_id + 1}/{args.chunks} ---")
+        if end_str:
+            cmd = ["python", "-m", "src.adsb.process_icao_chunk",
+                   "--chunk-id", str(chunk_id),
+                   "--total-chunks", str(args.chunks),
+                   "--start-date", start_str,
+                   "--end-date", end_str]
+        else:
+            cmd = ["python", "-m", "src.adsb.process_icao_chunk",
+                   "--chunk-id", str(chunk_id),
+                   "--total-chunks", str(args.chunks),
+                   "--date", start_str]
+        run_cmd(cmd, f"Process chunk {chunk_id}")
+    
+    # Step 3: Combine chunks to CSV
+    print("\n" + "=" * 60)
+    print("Step 3: Combine to CSV")
+    print("=" * 60)
+    
+    chunks_dir = "./data/output/adsb_chunks"
+    cmd = ["python", "-m", "src.adsb.combine_chunks_to_csv",
+           "--chunks-dir", chunks_dir]
+    
+    if end_str:
+        cmd.extend(["--start-date", start_str, "--end-date", end_str])
+    else:
+        cmd.extend(["--date", start_str])
+    
+    if args.skip_base:
+        cmd.append("--skip-base")
+    
+    run_cmd(cmd, "Combine chunks")
+    
+    print("\n" + "=" * 60)
+    print("Done!")
+    print("=" * 60)
+    
+    # Show output
+    output_dir = "./data/openairframes"
+    if end_str:
+        output_file = f"openairframes_adsb_{start_str}_{end_str}.csv"
+    else:
+        output_file = f"openairframes_adsb_{start_str}_{start_str}.csv"
+    
+    output_path = os.path.join(output_dir, output_file)
+    if os.path.exists(output_path):
+        size_mb = os.path.getsize(output_path) / (1024 * 1024)
+        print(f"Output: {output_path}")
+        print(f"Size: {size_mb:.1f} MB")
+
+
+if __name__ == "__main__":
+    main()

src/contributions/approve_submission.py

@@ -246,20 +246,6 @@ def process_submission(
     if schema_updated:
         schema_note = f"\n**Schema Updated:** Added new tags: `{', '.join(new_tags)}`\n"
     
-    # Truncate JSON preview to stay under GitHub's 65536 char body limit
-    max_json_preview = 50000
-    if len(content_json) > max_json_preview:
-        # Show first few entries as a preview
-        preview_entries = submissions[:10]
-        preview_json = json.dumps(preview_entries, indent=2, sort_keys=True)
-        json_section = (
-            f"### Submissions (showing 10 of {len(submissions)})\n"
-            f"```json\n{preview_json}\n```\n\n"
-            f"*Full submission ({len(submissions)} entries, {len(content_json):,} chars) is in the committed file.*"
-        )
-    else:
-        json_section = f"### Submissions\n```json\n{content_json}\n```"
-
     pr_body = f"""## Community Submission
 
 Adds {len(submissions)} submission(s) from @{author_username}.
@@ -271,7 +257,10 @@ Closes #{issue_number}
 
 ---
 
-{json_section}"""
+### Submissions
+```json
+{content_json}
+```"""
     
     pr = create_pull_request(
         title=f"Community submission: {filename}",

src/contributions/create_daily_community_release.py

@@ -24,7 +24,7 @@ def read_all_submissions(community_dir: Path) -> list[dict]:
     """Read all JSON submissions from the community directory."""
     all_submissions = []
     
-    for json_file in sorted(community_dir.glob("**/*.json")):
+    for json_file in sorted(community_dir.glob("*.json")):
         try:
             with open(json_file) as f:
                 data = json.load(f)

src/contributions/schema.py

@@ -36,52 +36,6 @@ def get_latest_schema_version() -> int:
     return max_version
 
 
-def _is_balanced_json(text: str) -> bool:
-    """
-    Check if JSON has balanced brackets/braces.
-    
-    This is a simple check to ensure we captured complete JSON.
-    Ignores brackets/braces inside strings.
-    
-    Args:
-        text: JSON text to check
-        
-    Returns:
-        True if balanced, False otherwise
-    """
-    in_string = False
-    escape = False
-    stack = []
-    pairs = {'[': ']', '{': '}'}
-    
-    for char in text:
-        if escape:
-            escape = False
-            continue
-            
-        if char == '\\':
-            escape = True
-            continue
-            
-        if char == '"' and not escape:
-            in_string = not in_string
-            continue
-            
-        if in_string:
-            continue
-            
-        if char in pairs:
-            stack.append(char)
-        elif char in pairs.values():
-            if not stack:
-                return False
-            if pairs[stack[-1]] != char:
-                return False
-            stack.pop()
-    
-    return len(stack) == 0 and not in_string
-
-
 def get_schema_path(version: int | None = None) -> Path:
     """
     Get path to a specific schema version, or latest if version is None.
@@ -208,14 +162,10 @@ def extract_json_from_issue_body(body: str) -> str | None:
         return match.group(1).strip()
     
     # Try: Raw JSON after "### Submission JSON" until next section or end
-    # Use greedy matching since we have a clear boundary (next ### or end)
-    pattern_raw = r"### Submission JSON\s*\n\s*([\[{][\s\S]*[\]}])(?=\s*\n###|\s*$)"
+    pattern_raw = r"### Submission JSON\s*\n\s*([\[{][\s\S]*?[\]}])(?=\n###|\n\n###|$)"
     match = re.search(pattern_raw, body)
     if match:
-        candidate = match.group(1).strip()
-        # Validate it's complete JSON by checking balanced brackets
-        if _is_balanced_json(candidate):
-            return candidate
+        return match.group(1).strip()
     
     # Try: Any JSON object/array in the body (fallback)
     pattern_any = r"([\[{][\s\S]*?[\]}])"
@@ -269,19 +219,7 @@ def parse_and_validate(json_str: str, schema: dict | None = None) -> tuple[list
     try:
         data = json.loads(json_str)
     except json.JSONDecodeError as e:
-        # Provide detailed error context
-        error_msg = f"Invalid JSON: {e}"
-        
-        # Show context around the error position
-        if hasattr(e, 'pos') and e.pos is not None:
-            start = max(0, e.pos - 50)
-            end = min(len(json_str), e.pos + 50)
-            context = json_str[start:end]
-            # Escape for readability
-            context_escaped = repr(context)
-            error_msg += f"\n\nContext around position {e.pos}: {context_escaped}"
-        
-        return None, [error_msg]
+        return None, [f"Invalid JSON: {e}"]
     
     errors = validate_submission(data, schema)
     return data, errors

src/create_daily_adsb_release.py

@@ -0,0 +1,84 @@
+from pathlib import Path
+from datetime import datetime, timezone, timedelta
+import sys
+
+import polars as pl
+
+# Add adsb directory to path
+sys.path.insert(0, str(Path(__file__).parent / "adsb")) # TODO: Fix this hacky path manipulation
+
+from adsb.compress_adsb_to_aircraft_data import (
+    load_historical_for_day,
+    concat_compressed_dfs,
+    get_latest_aircraft_adsb_csv_df,
+)
+
+if __name__ == '__main__':
+    # Get yesterday's date (data for the previous day)
+    day = datetime.now(timezone.utc) - timedelta(days=1)
+
+    # Find a day with complete data
+    max_attempts = 2  # Don't look back more than a week
+    for attempt in range(max_attempts):
+        date_str = day.strftime("%Y-%m-%d")
+        print(f"Processing ADS-B data for {date_str}")
+        
+        print("Loading new ADS-B data...")
+        df_new = load_historical_for_day(day)
+        if df_new.height == 0:
+            day = day - timedelta(days=1)
+            continue
+        max_time = df_new['time'].max()
+        if max_time is not None:
+            # Handle timezone
+            max_time_dt = max_time
+            if hasattr(max_time_dt, 'replace'):
+                max_time_dt = max_time_dt.replace(tzinfo=timezone.utc)
+            
+            end_of_day = day.replace(hour=23, minute=59, second=59, tzinfo=timezone.utc) - timedelta(minutes=5)
+            
+            # Convert polars datetime to python datetime if needed
+            if isinstance(max_time_dt, datetime):
+                if max_time_dt.replace(tzinfo=timezone.utc) >= end_of_day:
+                    break
+            else:
+                # Polars returns python datetime already
+                if max_time >= day.replace(hour=23, minute=54, second=59):
+                    break
+        
+        print(f"WARNING: Latest data time is {max_time}, which is more than 5 minutes before end of day.")
+        day = day - timedelta(days=1)
+    else:
+        raise RuntimeError(f"Could not find complete data in the last {max_attempts} days")
+
+    try:
+        # Get the latest release data
+        print("Downloading latest ADS-B release...")
+        df_base, start_date_str = get_latest_aircraft_adsb_csv_df()
+        # Combine with historical data
+        print("Combining with historical data...")
+        df_combined = concat_compressed_dfs(df_base, df_new)
+    except Exception as e:
+        print(f"Error downloading latest ADS-B release: {e}")
+        df_combined = df_new
+        start_date_str = date_str
+
+    # Sort by time for consistent ordering
+    df_combined = df_combined.sort('time')
+    
+    # Convert any list columns to strings for CSV compatibility
+    for col in df_combined.columns:
+        if df_combined[col].dtype == pl.List:
+            df_combined = df_combined.with_columns(
+                pl.col(col).list.join(",").alias(col)
+            )
+
+    # Save the result
+    OUT_ROOT = Path("data/openairframes")
+    OUT_ROOT.mkdir(parents=True, exist_ok=True)
+
+    output_file = OUT_ROOT / f"openairframes_adsb_{start_date_str}_{date_str}.csv.gz"
+    df_combined.write_csv(output_file)
+
+    print(f"Saved: {output_file}")
+    print(f"Total aircraft: {df_combined.height}")

src/get_latest_release.py

@@ -27,33 +27,6 @@ def _http_get_json(url: str, headers: dict[str, str]) -> dict:
     return json.loads(data.decode("utf-8"))
 
 
-def get_releases(repo: str = REPO, github_token: Optional[str] = None, per_page: int = 30) -> list[dict]:
-    """Get a list of releases from the repository."""
-    url = f"https://api.github.com/repos/{repo}/releases?per_page={per_page}"
-    headers = {
-        "Accept": "application/vnd.github+json",
-        "User-Agent": "openairframes-downloader/1.0",
-    }
-    if github_token:
-        headers["Authorization"] = f"Bearer {github_token}"
-
-    return _http_get_json(url, headers=headers)
-
-
-def get_release_assets_from_release_data(release_data: dict) -> list[ReleaseAsset]:
-    """Extract assets from a release data dictionary."""
-    assets = []
-    for a in release_data.get("assets", []):
-        assets.append(
-            ReleaseAsset(
-                name=a["name"],
-                download_url=a["browser_download_url"],
-                size=int(a.get("size", 0)),
-            )
-        )
-    return assets
-
-
 def get_latest_release_assets(repo: str = REPO, github_token: Optional[str] = None) -> list[ReleaseAsset]:
     url = f"https://api.github.com/repos/{repo}/releases/latest"
     headers = {
@@ -64,7 +37,16 @@ def get_latest_release_assets(repo: str = REPO, github_token: Optional[str] = No
         headers["Authorization"] = f"Bearer {github_token}"
 
     payload = _http_get_json(url, headers=headers)
-    return get_release_assets_from_release_data(payload)
+    assets = []
+    for a in payload.get("assets", []):
+        assets.append(
+            ReleaseAsset(
+                name=a["name"],
+                download_url=a["browser_download_url"],
+                size=int(a.get("size", 0)),
+            )
+        )
+    return assets
 
 
 def pick_asset(
@@ -173,8 +155,7 @@ def download_latest_aircraft_adsb_csv(
     repo: str = REPO,
 ) -> Path:
     """
-    Download the latest openairframes_adsb_*.csv file from GitHub releases.
-    If the latest release doesn't have the file, searches previous releases.
+    Download the latest openairframes_adsb_*.csv file from the latest GitHub release.
 
     Args:
         output_dir: Directory to save the downloaded file (default: "downloads")
@@ -185,69 +166,26 @@ def download_latest_aircraft_adsb_csv(
         Path to the downloaded file
     """
     output_dir = Path(output_dir)
-    
-    # Get multiple releases
-    releases = get_releases(repo, github_token=github_token, per_page=30)
-    
-    # Try each release until we find one with the matching asset
-    for release in releases:
-        assets = get_release_assets_from_release_data(release)
-        try:
-            asset = pick_asset(assets, name_regex=r"^openairframes_adsb_.*\.csv(\.gz)?$")
-            saved_to = download_asset(asset, output_dir / asset.name, github_token=github_token)
-            print(f"Downloaded: {asset.name} ({asset.size} bytes) -> {saved_to}")
-            return saved_to
-        except FileNotFoundError:
-            # This release doesn't have the matching asset, try the next one
-            continue
-    
-    raise FileNotFoundError(
-        f"No release in the last 30 releases has an asset matching 'openairframes_adsb_.*\\.csv(\\.gz)?$'"
-    )
+    assets = get_latest_release_assets(repo, github_token=github_token)
+    asset = pick_asset(assets, name_regex=r"^openairframes_adsb_.*\.csv(\.gz)?$")
+    saved_to = download_asset(asset, output_dir / asset.name, github_token=github_token)
+    print(f"Downloaded: {asset.name} ({asset.size} bytes) -> {saved_to}")
+    return saved_to
+
 
-import polars as pl
 def get_latest_aircraft_adsb_csv_df():
-    """Download and load the latest ADS-B CSV from GitHub releases.
-    
-    Returns:
-        tuple: (df, start_date, end_date) where dates are in YYYY-MM-DD format
-    """
-    import re
-    
     csv_path = download_latest_aircraft_adsb_csv()
-    df = pl.read_csv(csv_path, null_values=[""])
-    
-    # Parse time column: values like "2025-12-31T00:00:00.040" or "2025-05-11T15:15:50.540+0000"
-    # Try with timezone first (convert to naive), then without timezone
-    df = df.with_columns(
-        pl.col("time").str.strptime(pl.Datetime("ms"), "%Y-%m-%dT%H:%M:%S%.f%z", strict=False)
-            .dt.replace_time_zone(None)  # Convert to naive datetime first
-            .fill_null(pl.col("time").str.strptime(pl.Datetime("ms"), "%Y-%m-%dT%H:%M:%S%.f", strict=False))
-    )
-
-    # Cast dbFlags and year to strings to match the schema used in compress functions
-    for col in ['dbFlags', 'year']:
-        if col in df.columns:
-            df = df.with_columns(pl.col(col).cast(pl.Utf8))
-    
-    # Fill nulls with empty strings for string columns
-    for col in df.columns:
-        if df[col].dtype == pl.Utf8:
-            df = df.with_columns(pl.col(col).fill_null(""))
-    
-    # Extract start and end dates from filename pattern: openairframes_adsb_{start_date}_{end_date}.csv[.gz]
-    match = re.search(r"openairframes_adsb_(\d{4}-\d{2}-\d{2})_(\d{4}-\d{2}-\d{2})\.csv", str(csv_path))
+    import pandas as pd
+    df = pd.read_csv(csv_path)
+    df = df.fillna("")
+    # Extract start date from filename pattern: openairframes_adsb_{start_date}_{end_date}.csv[.gz]
+    match = re.search(r"openairframes_adsb_(\d{4}-\d{2}-\d{2})_", str(csv_path))
     if not match:
-        raise ValueError(f"Could not extract dates from filename: {csv_path.name}")
+        raise ValueError(f"Could not extract date from filename: {csv_path.name}")
     
-    start_date = match.group(1)
-    end_date = match.group(2)
-    print(df.columns)
-    print(df.dtypes)
-    return df, start_date, end_date
-
+    date_str = match.group(1)
+    return df, date_str
 
 
 if __name__ == "__main__":
     download_latest_aircraft_csv()
-    download_latest_aircraft_adsb_csv()