delete parquet chunck after load to not use so much space for big historical run

.github/workflows/historical-adsb.yaml

@@ -188,17 +188,19 @@ jobs:
 
       - 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 | head -50 || echo "No files found"
-          echo "=== Looking for parquet files ==="
-          find . -name "*.parquet" 2>/dev/null | head -20 || echo "No parquet files found"
+          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 "$END_DATE" --skip-base
+          python -m src.adsb.combine_chunks_to_csv --chunks-dir data/output/adsb_chunks --start-date "$START_DATE" --end-date "$END_DATE" --skip-base --stream
           ls -lah data/planequery_aircraft/
 
       - name: Upload final artifact

src/adsb/combine_chunks_to_csv.py

@@ -36,8 +36,13 @@ def get_target_day() -> datetime:
     return datetime.utcnow() - timedelta(days=1)
 
 
-def process_single_chunk(chunk_path: str) -> pl.DataFrame:
-    """Load and compress a single chunk parquet file."""
+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
@@ -45,6 +50,14 @@ def process_single_chunk(chunk_path: str) -> pl.DataFrame:
     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")
@@ -156,6 +169,7 @@ def main():
     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
@@ -190,9 +204,10 @@ def main():
     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)
+        compressed = process_single_chunk(chunk_path, delete_after_load=args.stream)
         compressed_chunks.append(compressed)
         gc.collect()