Add Polars-based compile script for speed comparison

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

compile_polars.py

@@ -0,0 +1,106 @@
+# /// script
+# requires-python = ">=3.11"
+# dependencies = [
+#     "huggingface-hub>=1.7",
+#     "polars",
+#     "python-dotenv",
+#     "structlog",
+#     "datasets",
+# ]
+# ///
+"""Quick test: sync bucket + Polars for compile speed comparison."""
+
+import os
+import shutil
+import tempfile
+import time
+
+import polars as pl
+import structlog
+from datasets import Dataset
+from dotenv import load_dotenv
+from huggingface_hub import login, sync_bucket
+
+load_dotenv(override=True)
+
+HF_TOKEN = os.getenv("HF_TOKEN")
+assert HF_TOKEN is not None
+
+BUCKET_ID = os.getenv("BUCKET_ID", "librarian-bots/card-store")
+OUTPUT_REPO = os.getenv(
+    "OUTPUT_REPO", "librarian-bots/dataset_cards_with_metadata_test"
+)
+DRY_RUN = os.getenv("DRY_RUN", "false").lower() == "true"
+
+structlog.configure(
+    processors=[
+        structlog.stdlib.add_log_level,
+        structlog.processors.TimeStamper(fmt="iso"),
+        structlog.dev.ConsoleRenderer(),
+    ],
+    wrapper_class=structlog.make_filtering_bound_logger(20),
+    context_class=dict,
+    logger_factory=structlog.PrintLoggerFactory(),
+    cache_logger_on_first_use=True,
+)
+logger = structlog.get_logger()
+
+login(token=HF_TOKEN)
+
+
+def main():
+    t0 = time.time()
+
+    # Step 1: Sync bucket
+    local_dir = tempfile.mkdtemp(prefix="card-compile-")
+    bucket_path = f"hf://buckets/{BUCKET_ID}/datasets"
+    logger.info("Syncing bucket", bucket=bucket_path)
+    sync_bucket(source=bucket_path, dest=local_dir, token=HF_TOKEN)
+    t1 = time.time()
+    logger.info("Sync done", seconds=round(t1 - t0, 1))
+
+    # Step 2: Read all JSON with Polars
+    json_files = [
+        os.path.join(local_dir, f) for f in os.listdir(local_dir) if f.endswith(".json")
+    ]
+    logger.info("Reading JSON files", count=len(json_files))
+
+    # Read all files into a single DataFrame
+    dfs = []
+    for f in json_files:
+        try:
+            dfs.append(pl.read_json(f))
+        except Exception:
+            pass
+    df = pl.concat(dfs)
+    t2 = time.time()
+    logger.info("Polars read done", rows=len(df), seconds=round(t2 - t1, 1))
+
+    # Step 3: Filter
+    df = df.filter(pl.col("card").is_not_null())
+    df = df.filter(pl.col("card").str.len_bytes() > 30)
+    df = df.filter(~pl.col("card").str.contains("<h1>429</h1>"))
+    df = df.filter(~pl.col("card").str.contains("Entry not found"))
+
+    # Drop fetched_at
+    if "fetched_at" in df.columns:
+        df = df.drop("fetched_at")
+
+    t3 = time.time()
+    logger.info("Filter done", rows=len(df), seconds=round(t3 - t2, 1))
+
+    # Step 4: Push
+    if DRY_RUN:
+        logger.info("[DRY RUN]", rows=len(df), columns=df.columns)
+    else:
+        logger.info("Pushing to Hub", repo=OUTPUT_REPO)
+        ds = Dataset.from_polars(df)
+        ds.push_to_hub(OUTPUT_REPO, token=HF_TOKEN, max_shard_size="750MB")
+        t4 = time.time()
+        logger.info("Done", rows=len(df), total_seconds=round(t4 - t0, 1))
+
+    shutil.rmtree(local_dir, ignore_errors=True)
+
+
+if __name__ == "__main__":
+    main()