Update README.md

README.md

@@ -1,5 +1,140 @@
 ---
-license: cc-by-4.0
+pretty_name: Spotify Public Playlists (Oct 2024)
+dataset_summary: >-
+  CSV export of 91,906 Spotify public playlists captured from a PostgreSQL dump
+  in October 2024, including playlist metadata, owner info, follower counts,
+  cover images, JSON blobs with the raw Spotify payload, and historical
+  snapshots.
 language:
-- en
----
+  - en
+license: cc-by-4.0
+tags:
+  - music
+  - spotify
+  - playlists
+  - tabular-data
+size_categories:
+  - 10K<n<100K
+task_categories:
+  - tabular-classification
+---
+
+# Spotify Public Playlists (October 2024)
+
+This repository contains an open dataset with **91,906** Spotify playlists that
+were present in a PostgreSQL backup captured in **October 2024**. Each row
+represents a playlist and includes high-level metadata (name, owner, follower
+count), cover image URLs, JSON payloads with the full Spotify response, and
+various bookkeeping columns (status, popularity histograms, history, etc.).
+
+The dataset ships as a ready-to-use CSV (`playlists.csv`, ~9 GB) along with the
+raw dump (`playlists_full.sql`, 27 GB) and a Python extractor script
+(`export_playlists_to_csv.py`) so anyone can regenerate the CSV from source.
+
+> ⚠️ **Attribution + ToS**: This dataset aggregates Spotify metadata originally
+> provided via their APIs. Redistribution is offered under CC BY 4.0, but you are
+> responsible for complying with Spotify's terms of service and ensuring any
+> downstream use respects user privacy.
+
+## Files
+
+| Path | Description |
+| --- | --- |
+| `playlists_full.sql` | Original PostgreSQL dump (27 GB) containing `INSERT` statements for every table. |
+| `export_playlists_to_csv.py` | Streaming parser that extracts only `public.playlists` rows into CSV with progress reporting and batching. |
+| `playlists.csv` | Final 91,906-row CSV (17 columns, ~9 GB) with one playlist per row. |
+| `LICENSE` | CC BY 4.0 license text covering this dataset and helper scripts. |
+
+## Schema
+
+| Column | Type | Notes |
+| --- | --- | --- |
+| `id` | integer | Autoincrement primary key from the source database. |
+| `playlist_id` | string | Spotify playlist ID. |
+| `name` | string | Playlist display name. |
+| `description` | string | Spotify-provided description (HTML/emoji possible). |
+| `followers_count` | integer | Followers when the snapshot was taken. |
+| `owner_id` | string | Spotify user ID of the owner. |
+| `owner_name` | string | Display name of the owner. |
+| `is_public` | boolean | Whether the playlist was publicly visible. |
+| `tracks_total` | integer | Number of tracks reported by Spotify. |
+| `snapshot_id` | string | Snapshot token returned by Spotify for change tracking. |
+| `image_url` | string | Primary cover image URL. |
+| `raw_data` | JSON (string) | Full Spotify playlist payload as JSON (can exceed 1 MB). |
+| `created_at` | timestamp | When this playlist was inserted into the source DB. |
+| `updated_at` | timestamp | Last time the playlist row was updated. |
+| `followers_history` | JSON (string) | Historical follower counts (if tracked). |
+| `status` | string | ETL status flag from the upstream system. |
+| `popularity_distribution` | JSON (string) | Histogram of track popularity buckets. |
+
+## Regenerating the CSV
+
+```bash
+python3 export_playlists_to_csv.py \
+  --batch-size 2000 \
+  --progress-interval 0.5
+```
+
+Arguments:
+
+- `--batch-size` (default 2000): number of rows buffered before writing to CSV.
+- `--progress-interval` (default 0.25 seconds): how often to refresh the
+  progress line.
+
+The script streams the SQL file, stops after the playlists table, and prints
+live progress (`rows`, `% of dump read`, and GiB consumed). It sets
+`csv.field_size_limit` to accommodate multi-megabyte JSON blobs.
+
+## Loading the Dataset
+
+Once the CSV is available (locally or via Hugging Face), you can load it with
+your favorite data tooling:
+
+```python
+import pandas as pd
+
+df = pd.read_csv(
+    "playlists.csv",
+    dtype={"playlist_id": str, "owner_id": str},
+    converters={"raw_data": lambda x: x},  # keep JSON as raw strings
+)
+print(len(df))  # 91906
+```
+
+For streaming workflows (due to the 9 GB size), consider `pyarrow.dataset` or
+chunked `pandas.read_csv(..., chunksize=10_000)`.
+
+## Publishing to Hugging Face
+
+1. **Create dataset repo**: `huggingface-cli repo create <namespace>/spotify-playlists-2024 --type dataset`.
+2. **Clone with Git LFS**: `GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/datasets/<namespace>/spotify-playlists-2024` then `cd` into it and run `git lfs install`.
+3. **Copy assets**: Bring over `playlists.csv`, `export_playlists_to_csv.py`, `README.md`, and `LICENSE`.
+4. **Track large files**: `git lfs track "*.csv" "*.sql"`. Commit the updated `.gitattributes`.
+5. **Push**: `git add . && git commit -m "Add Spotify playlists dataset" && git push`.
+6. **Add metadata**: The `README.md` you are reading doubles as the Hugging Face dataset card, so it will render on the hub automatically.
+
+Optional: Upload `playlists_full.sql` as a release artifact or keep it in LFS if
+you want to allow full reproducibility for others.
+
+## License
+
+Creative Commons Attribution 4.0 International (CC BY 4.0). See `LICENSE` for
+the full text. When using the dataset, credit “Spotify Public Playlists (Oct 2024)”
+and include a link back to the Hugging Face dataset page or this repository.
+
+## Citation
+
+```
+@dataset{spotify_public_playlists_oct2024,
+  title        = {Spotify Public Playlists (October 2024)},
+  author       = {Max and contributors},
+  year         = {2024},
+  howpublished = {Hugging Face Datasets},
+  note         = {CC-BY 4.0}
+}
+```
+
+## Contact
+
+Questions or takedown requests? Open an issue or reach out via the contact info
+on my Hugging Face profile.