| --- |
| license: cc-by-sa-4.0 |
| tags: |
| - chemistry |
| - reactions |
| - cheminformatics |
| pretty_name: The Open Reaction Database |
| configs: |
| - config_name: default |
| default: true |
| data_files: |
| - data/00/ord_dataset-00005539a1e04c809a9a78647bea649c.parquet |
| - data/03/ord_dataset-0316886541d9435489859c2ad7edc863.parquet |
| - data/0c/ord_dataset-0c75d67751634f0594b24b9f498b77c2.parquet |
| - data/10/ord_dataset-10b940e7982c4622b1e1ac879394aba6.parquet |
| - data/17/ord_dataset-172039a759a440219a68af62d203b79b.parquet |
| - data/1d/ord_dataset-1d3a1a6fb70d46c084602d0688967afc.parquet |
| - data/1e/ord_dataset-1ec2807f02fa4beda27d2a81b86eb843.parquet |
| - data/20/ord_dataset-2038a3c967db4a32a8fbce288437e929.parquet |
| - data/2b/ord_dataset-2be11f57f3304e678ea8469baf6dd1bc.parquet |
| - data/2b/ord_dataset-2be30f5d8dcd471aa6ad410bdee05902.parquet |
| - data/31/ord_dataset-31989f1b2b9d4885b1dd2d9982da4517.parquet |
| - data/35/ord_dataset-35a5a513f1dd44a3a97c88da99f81a00.parquet |
| - data/3b/ord_dataset-3b5db90e337942ea886b8f5bc5e3aa72.parquet |
| - data/3b/ord_dataset-3b8a2ef300e145468579027f206a3ac8.parquet |
| - data/46/ord_dataset-46ff9a32d9e04016b9380b1b1ef949c3.parquet |
| - data/47/ord_dataset-47eaacc46c3a4487bbdf99adb1a15e41.parquet |
| - data/4d/ord_dataset-4d431564f3ef4e9c91d8da5836f4eae6.parquet |
| - data/52/ord_dataset-52bd3b0ec72c443aab113bcea09bf3f4.parquet |
| - data/54/ord_dataset-5415f83da68e4797856ddd330dda5c61.parquet |
| - data/55/ord_dataset-5540e162c09f4c04905ddc8ba9c931c6.parquet |
| - data/55/ord_dataset-55de08a995554f558c25fc43eac62359.parquet |
| - data/5c/ord_dataset-5c9a10329a8a48968d18879a48bb8ab2.parquet |
| - data/5e/ord_dataset-5e8318f0dda04b398c14f4c3adfeb32c.parquet |
| - data/5e/ord_dataset-5eb7f2689f4a42eba63ad9e37e49a5cd.parquet |
| - data/67/ord_dataset-675eddcaa6674ce3ae61e79bbc1e1c08.parquet |
| - data/68/ord_dataset-68cb8b4b2b384e3d85b5b1efae58b203.parquet |
| - data/6a/ord_dataset-6a0bfcdf53a64c07987822162ae591e2.parquet |
| - data/7a/ord_dataset-7acd6ad2bf4d4cff841cad008ab726d5.parquet |
| - data/7d/ord_dataset-7d8f5fd922d4497d91cb81489b052746.parquet |
| - data/80/ord_dataset-805ad863feef48579d95d86a728035f4.parquet |
| - data/89/ord_dataset-89b083710e2d441aa0040c361d63359f.parquet |
| - data/8d/ord_dataset-8d1e28ec1f6b4ee8ad372e0b5ed7a62e.parquet |
| - data/99/ord_dataset-99c23cf435dc42f1af884053bc8b11c7.parquet |
| - data/9b/ord_dataset-9b8aa9a7835143ef8ce3f70abfab7545.parquet |
| - data/a1/ord_dataset-a12fa15d036d489c971b0b514caeae52.parquet |
| - data/ac/ord_dataset-ac78456835404910b3a4c840248b6ac9.parquet |
| - data/b4/ord_dataset-b440f8c90b6343189093770060fc4098.parquet |
| - data/bc/ord_dataset-bc349c19b4384756a2aee8aa525b6c2a.parquet |
| - data/c5/ord_dataset-c5b00523487a4211a194160edf45e9ab.parquet |
| - data/c7/ord_dataset-c703268ea43a4c7e802c6048b6166b34.parquet |
| - data/cb/ord_dataset-cbcc4048add7468e850b6ec42549c70d.parquet |
| - data/ce/ord_dataset-ce5045aceb214cfc8bfd0ef3031e2737.parquet |
| - data/d2/ord_dataset-d26118acda314269becc35db5c22dc59.parquet |
| - data/d3/ord_dataset-d319c2a22ecf4ce59db1a18ae71d529c.parquet |
| - data/d9/ord_dataset-d92976309c3a48a3a64a4cf5e7048086.parquet |
| - data/dc/ord_dataset-dc0249930af34d17a3c76881a762aebf.parquet |
| - data/ee/ord_dataset-eeba974d3c284aed86d1c1d442260a1e.parquet |
| - data/fc/ord_dataset-fc83743b978f4deea7d6856deacbfe53.parquet |
| - data/fe/ord_dataset-feaf1b793c6d408aaec1cac7cc3ceadc.parquet |
| - config_name: uspto-grants |
| data_files: |
| - data/11/ord_dataset-1158e351757f315b93cbcbe7bc55f38e.parquet |
| - config_name: uspto-mit |
| data_files: |
| - split: train |
| path: data/e7/ord_dataset-e7830cd6b11158b43994ccfb5ee9acb3.parquet |
| - split: validation |
| path: data/54/ord_dataset-5481550056a14935b76e031fb94b88be.parquet |
| - split: test |
| path: data/48/ord_dataset-488402f6ec0d441ca2f7d6fabea7c220.parquet |
| - config_name: ord_dataset-00005539a1e04c809a9a78647bea649c |
| data_files: |
| - data/00/ord_dataset-00005539a1e04c809a9a78647bea649c.parquet |
| - config_name: ord_dataset-0316886541d9435489859c2ad7edc863 |
| data_files: |
| - data/03/ord_dataset-0316886541d9435489859c2ad7edc863.parquet |
| - config_name: ord_dataset-0c75d67751634f0594b24b9f498b77c2 |
| data_files: |
| - data/0c/ord_dataset-0c75d67751634f0594b24b9f498b77c2.parquet |
| - config_name: ord_dataset-10b940e7982c4622b1e1ac879394aba6 |
| data_files: |
| - data/10/ord_dataset-10b940e7982c4622b1e1ac879394aba6.parquet |
| - config_name: ord_dataset-1158e351757f315b93cbcbe7bc55f38e |
| data_files: |
| - data/11/ord_dataset-1158e351757f315b93cbcbe7bc55f38e.parquet |
| - config_name: ord_dataset-172039a759a440219a68af62d203b79b |
| data_files: |
| - data/17/ord_dataset-172039a759a440219a68af62d203b79b.parquet |
| - config_name: ord_dataset-1d3a1a6fb70d46c084602d0688967afc |
| data_files: |
| - data/1d/ord_dataset-1d3a1a6fb70d46c084602d0688967afc.parquet |
| - config_name: ord_dataset-1ec2807f02fa4beda27d2a81b86eb843 |
| data_files: |
| - data/1e/ord_dataset-1ec2807f02fa4beda27d2a81b86eb843.parquet |
| - config_name: ord_dataset-2038a3c967db4a32a8fbce288437e929 |
| data_files: |
| - data/20/ord_dataset-2038a3c967db4a32a8fbce288437e929.parquet |
| - config_name: ord_dataset-2be11f57f3304e678ea8469baf6dd1bc |
| data_files: |
| - data/2b/ord_dataset-2be11f57f3304e678ea8469baf6dd1bc.parquet |
| - config_name: ord_dataset-2be30f5d8dcd471aa6ad410bdee05902 |
| data_files: |
| - data/2b/ord_dataset-2be30f5d8dcd471aa6ad410bdee05902.parquet |
| - config_name: ord_dataset-31989f1b2b9d4885b1dd2d9982da4517 |
| data_files: |
| - data/31/ord_dataset-31989f1b2b9d4885b1dd2d9982da4517.parquet |
| - config_name: ord_dataset-35a5a513f1dd44a3a97c88da99f81a00 |
| data_files: |
| - data/35/ord_dataset-35a5a513f1dd44a3a97c88da99f81a00.parquet |
| - config_name: ord_dataset-3b5db90e337942ea886b8f5bc5e3aa72 |
| data_files: |
| - data/3b/ord_dataset-3b5db90e337942ea886b8f5bc5e3aa72.parquet |
| - config_name: ord_dataset-3b8a2ef300e145468579027f206a3ac8 |
| data_files: |
| - data/3b/ord_dataset-3b8a2ef300e145468579027f206a3ac8.parquet |
| - config_name: ord_dataset-46ff9a32d9e04016b9380b1b1ef949c3 |
| data_files: |
| - data/46/ord_dataset-46ff9a32d9e04016b9380b1b1ef949c3.parquet |
| - config_name: ord_dataset-47eaacc46c3a4487bbdf99adb1a15e41 |
| data_files: |
| - data/47/ord_dataset-47eaacc46c3a4487bbdf99adb1a15e41.parquet |
| - config_name: ord_dataset-488402f6ec0d441ca2f7d6fabea7c220 |
| data_files: |
| - data/48/ord_dataset-488402f6ec0d441ca2f7d6fabea7c220.parquet |
| - config_name: ord_dataset-4d431564f3ef4e9c91d8da5836f4eae6 |
| data_files: |
| - data/4d/ord_dataset-4d431564f3ef4e9c91d8da5836f4eae6.parquet |
| - config_name: ord_dataset-52bd3b0ec72c443aab113bcea09bf3f4 |
| data_files: |
| - data/52/ord_dataset-52bd3b0ec72c443aab113bcea09bf3f4.parquet |
| - config_name: ord_dataset-5415f83da68e4797856ddd330dda5c61 |
| data_files: |
| - data/54/ord_dataset-5415f83da68e4797856ddd330dda5c61.parquet |
| - config_name: ord_dataset-5481550056a14935b76e031fb94b88be |
| data_files: |
| - data/54/ord_dataset-5481550056a14935b76e031fb94b88be.parquet |
| - config_name: ord_dataset-5540e162c09f4c04905ddc8ba9c931c6 |
| data_files: |
| - data/55/ord_dataset-5540e162c09f4c04905ddc8ba9c931c6.parquet |
| - config_name: ord_dataset-55de08a995554f558c25fc43eac62359 |
| data_files: |
| - data/55/ord_dataset-55de08a995554f558c25fc43eac62359.parquet |
| - config_name: ord_dataset-5c9a10329a8a48968d18879a48bb8ab2 |
| data_files: |
| - data/5c/ord_dataset-5c9a10329a8a48968d18879a48bb8ab2.parquet |
| - config_name: ord_dataset-5e8318f0dda04b398c14f4c3adfeb32c |
| data_files: |
| - data/5e/ord_dataset-5e8318f0dda04b398c14f4c3adfeb32c.parquet |
| - config_name: ord_dataset-5eb7f2689f4a42eba63ad9e37e49a5cd |
| data_files: |
| - data/5e/ord_dataset-5eb7f2689f4a42eba63ad9e37e49a5cd.parquet |
| - config_name: ord_dataset-675eddcaa6674ce3ae61e79bbc1e1c08 |
| data_files: |
| - data/67/ord_dataset-675eddcaa6674ce3ae61e79bbc1e1c08.parquet |
| - config_name: ord_dataset-68cb8b4b2b384e3d85b5b1efae58b203 |
| data_files: |
| - data/68/ord_dataset-68cb8b4b2b384e3d85b5b1efae58b203.parquet |
| - config_name: ord_dataset-6a0bfcdf53a64c07987822162ae591e2 |
| data_files: |
| - data/6a/ord_dataset-6a0bfcdf53a64c07987822162ae591e2.parquet |
| - config_name: ord_dataset-7acd6ad2bf4d4cff841cad008ab726d5 |
| data_files: |
| - data/7a/ord_dataset-7acd6ad2bf4d4cff841cad008ab726d5.parquet |
| - config_name: ord_dataset-7d8f5fd922d4497d91cb81489b052746 |
| data_files: |
| - data/7d/ord_dataset-7d8f5fd922d4497d91cb81489b052746.parquet |
| - config_name: ord_dataset-805ad863feef48579d95d86a728035f4 |
| data_files: |
| - data/80/ord_dataset-805ad863feef48579d95d86a728035f4.parquet |
| - config_name: ord_dataset-89b083710e2d441aa0040c361d63359f |
| data_files: |
| - data/89/ord_dataset-89b083710e2d441aa0040c361d63359f.parquet |
| - config_name: ord_dataset-8d1e28ec1f6b4ee8ad372e0b5ed7a62e |
| data_files: |
| - data/8d/ord_dataset-8d1e28ec1f6b4ee8ad372e0b5ed7a62e.parquet |
| - config_name: ord_dataset-99c23cf435dc42f1af884053bc8b11c7 |
| data_files: |
| - data/99/ord_dataset-99c23cf435dc42f1af884053bc8b11c7.parquet |
| - config_name: ord_dataset-9b8aa9a7835143ef8ce3f70abfab7545 |
| data_files: |
| - data/9b/ord_dataset-9b8aa9a7835143ef8ce3f70abfab7545.parquet |
| - config_name: ord_dataset-a12fa15d036d489c971b0b514caeae52 |
| data_files: |
| - data/a1/ord_dataset-a12fa15d036d489c971b0b514caeae52.parquet |
| - config_name: ord_dataset-ac78456835404910b3a4c840248b6ac9 |
| data_files: |
| - data/ac/ord_dataset-ac78456835404910b3a4c840248b6ac9.parquet |
| - config_name: ord_dataset-b440f8c90b6343189093770060fc4098 |
| data_files: |
| - data/b4/ord_dataset-b440f8c90b6343189093770060fc4098.parquet |
| - config_name: ord_dataset-bc349c19b4384756a2aee8aa525b6c2a |
| data_files: |
| - data/bc/ord_dataset-bc349c19b4384756a2aee8aa525b6c2a.parquet |
| - config_name: ord_dataset-c5b00523487a4211a194160edf45e9ab |
| data_files: |
| - data/c5/ord_dataset-c5b00523487a4211a194160edf45e9ab.parquet |
| - config_name: ord_dataset-c703268ea43a4c7e802c6048b6166b34 |
| data_files: |
| - data/c7/ord_dataset-c703268ea43a4c7e802c6048b6166b34.parquet |
| - config_name: ord_dataset-cbcc4048add7468e850b6ec42549c70d |
| data_files: |
| - data/cb/ord_dataset-cbcc4048add7468e850b6ec42549c70d.parquet |
| - config_name: ord_dataset-ce5045aceb214cfc8bfd0ef3031e2737 |
| data_files: |
| - data/ce/ord_dataset-ce5045aceb214cfc8bfd0ef3031e2737.parquet |
| - config_name: ord_dataset-d26118acda314269becc35db5c22dc59 |
| data_files: |
| - data/d2/ord_dataset-d26118acda314269becc35db5c22dc59.parquet |
| - config_name: ord_dataset-d319c2a22ecf4ce59db1a18ae71d529c |
| data_files: |
| - data/d3/ord_dataset-d319c2a22ecf4ce59db1a18ae71d529c.parquet |
| - config_name: ord_dataset-d92976309c3a48a3a64a4cf5e7048086 |
| data_files: |
| - data/d9/ord_dataset-d92976309c3a48a3a64a4cf5e7048086.parquet |
| - config_name: ord_dataset-dc0249930af34d17a3c76881a762aebf |
| data_files: |
| - data/dc/ord_dataset-dc0249930af34d17a3c76881a762aebf.parquet |
| - config_name: ord_dataset-e7830cd6b11158b43994ccfb5ee9acb3 |
| data_files: |
| - data/e7/ord_dataset-e7830cd6b11158b43994ccfb5ee9acb3.parquet |
| - config_name: ord_dataset-eeba974d3c284aed86d1c1d442260a1e |
| data_files: |
| - data/ee/ord_dataset-eeba974d3c284aed86d1c1d442260a1e.parquet |
| - config_name: ord_dataset-fc83743b978f4deea7d6856deacbfe53 |
| data_files: |
| - data/fc/ord_dataset-fc83743b978f4deea7d6856deacbfe53.parquet |
| - config_name: ord_dataset-feaf1b793c6d408aaec1cac7cc3ceadc |
| data_files: |
| - data/fe/ord_dataset-feaf1b793c6d408aaec1cac7cc3ceadc.parquet |
| --- |
| |
| # ord-data |
|
|
|  |
| [](https://zenodo.org/badge/latestdoi/283813042) |
|
|
| ## Getting the Data |
|
|
| The datasets live under [`data/`](data) and are stored with |
| [Git LFS](https://git-lfs.com/). LFS reads are redirected to the |
| [Hugging Face mirror](https://huggingface.co/datasets/open-reaction-database/ord-data) |
| via [`.lfsconfig`](.lfsconfig), so dataset objects are fetched from Hugging |
| Face's CDN rather than from GitHub's shared (and limited) LFS bandwidth. This is |
| automatic — you do not need to configure anything. |
|
|
| ### Option 1: Clone the repository |
|
|
| ```bash |
| git clone https://github.com/open-reaction-database/ord-data.git |
| ``` |
|
|
| With [Git LFS](https://git-lfs.com/) installed, this pulls every dataset object |
| from the Hugging Face mirror and gives you the full Git history with the data in |
| place. |
|
|
| ### Option 2: Download only the data (a subset, or without Git history) |
|
|
| ```bash |
| pip install huggingface_hub |
| python scripts/download_from_huggingface.py |
| ``` |
|
|
| (If you use [uv](https://docs.astral.sh/uv/), `uv run scripts/download_from_huggingface.py` |
| installs the dependencies for you.) |
|
|
| The script mirrors the `data/` directory from the Hugging Face dataset into your |
| local checkout. Pass `--allow-pattern 'data/4d/*.parquet'` (repeatable) to download |
| only a subset, or `--output-dir <path>` to write somewhere other than the |
| repository root. To skip LFS entirely during the clone and fetch the data |
| afterward: |
|
|
| ```bash |
| GIT_LFS_SKIP_SMUDGE=1 git clone https://github.com/open-reaction-database/ord-data.git |
| cd ord-data |
| python scripts/download_from_huggingface.py |
| ``` |
|
|
| You can also browse and download datasets directly from the |
| [Hugging Face dataset page](https://huggingface.co/datasets/open-reaction-database/ord-data). |
|
|
| For how this LFS / Hugging Face mirror setup works (and what it means for |
| contributors), see |
| [Git LFS and the Hugging Face mirror](#git-lfs-and-the-hugging-face-mirror) |
| below. |
|
|
| ## Data Manipulation |
|
|
| The `ord-data` repository contains the Open Reaction Database (ORD) as Parquet files stored in the |
| [`data`](data) directory. Each file holds one dataset: the reactions are serialized Protobuf |
| messages (one per row) and the dataset name/description/ID travel in the file metadata, so those |
| scalars and every reaction survive a round trip through `ord_schema`. (`Dataset.reaction_ids`, |
| which names reactions stored outside the dataset, is not persisted — the `reaction_id` column is |
| the source of truth — and a dataset with no reactions cannot be written at all.) The user can |
| convert the data into human readable text format, *.pbtxt. |
| |
| ```python |
| # import requirements |
| from ord_schema.datasets import load_dataset |
| from ord_schema.message_helpers import save_message |
| |
| # load the parquet file as a Dataset proto |
| # (as_dataset materializes it; the default is a streaming DatasetView) |
| dataset = load_dataset("input_fname.parquet", as_dataset=True) |
| # save the ord file as human readable text |
| save_message(dataset, "output_fname.pbtxt") |
| ``` |
| |
| `datasets.load_dataset` returns a `DatasetView` for Parquet, which takes the |
| dataset's scalars and row count from the file footer and reads reactions on |
| demand. (Note the name collision: `ord_schema.parquet.load_dataset` always |
| materializes, which on the largest file is a silent 1.8M-reaction load rather |
| than an error.) Iteration, `len`, indexing, and slicing over `reactions` behave |
| like a list, so read-only code needs no change — and the USPTO grants file, |
| 1.8M reactions in 1.1 GB, opens in milliseconds rather than deserializing up |
| front. Two places the resemblance stops: indexing returns a freshly |
| deserialized copy rather than a live sub-message, so mutating it changes |
| nothing on disk, and `reactions` never compares equal to a list — compare |
| against `list(view.reactions)`. Pass `as_dataset=True` (above) when you need the |
| protobuf surface: serializing or mutating the dataset as a whole. If you already |
| hold a view, `to_proto()` does the same thing. |
| |
| Beyond plain iteration, a view can look a reaction up by ID with `get_reaction`, |
| yield IDs alone with `iter_reaction_ids`, and read one row group at a time with |
| `iter_reactions(row_group=...)`, which yields `(reaction_id, Reaction)` pairs |
| rather than bare reactions. Row groups are the unit of parallelism for fanning |
| out over a large file; `view.num_row_groups` gives the upper bound, and an index |
| outside it raises `IndexError`. |
| |
| We can also convert ORD data into JSON format. |
| |
| ```python |
| # import requirements |
| import json |
| |
| from ord_schema.datasets import load_dataset |
| from google.protobuf.json_format import MessageToJson |
| |
| input_fname = "sample_file.parquet" |
| dataset = load_dataset(input_fname) |
| |
| # take one reaction message from the dataset for example |
| rxn = dataset.reactions[0] |
| rxn_json = json.loads( |
| MessageToJson( |
| message=rxn, |
| always_print_fields_with_no_presence=False, |
| preserving_proto_field_name=True, |
| indent=2, |
| sort_keys=False, |
| use_integers_for_enums=False, |
| descriptor_pool=None, |
| float_precision=None, |
| ensure_ascii=True, |
| ) |
| ) |
| |
| print( |
| f"We have converted the {input_fname} to JSON format shown as below, \n{rxn_json}" |
| ) |
| ``` |
| |
| ## Retired dataset IDs |
| |
| Two datasets are consolidations of shards that were once published separately, and a consolidated |
| output carries a new ID derived from the sorted IDs of its sources: |
| |
| | Dataset | ID | Sources | |
| | --- | --- | --- | |
| | `uspto-grants` | `ord_dataset-1158e351757f315b93cbcbe7bc55f38e` | 489 monthly `uspto-grants-YYYY_MM` datasets | |
| | `Training data from https://doi.org/10.1039/C8SC04228D` | `ord_dataset-e7830cd6b11158b43994ccfb5ee9acb3` | 10 `(N/10)` shards | |
| |
| [`retired_datasets.csv`](retired_datasets.csv) maps each of the 499 retired IDs to the dataset that |
| replaced it. Every other dataset kept its ID. |
| |
| Per-reaction patent provenance for the USPTO data is on `Reaction.provenance.patent`, so the |
| monthly bucket a reaction came from is recoverable from the consolidated dataset. |
| |
| ## Git LFS and the Hugging Face mirror |
| |
| Dataset files under [`data/`](data) are stored with Git LFS. Clone and fork |
| traffic was dominating GitHub's shared LFS bandwidth quota, so the repository is |
| configured to keep that traffic off GitHub while leaving GitHub authoritative |
| for the data: |
| |
| - **Reads come from Hugging Face.** [`.lfsconfig`](.lfsconfig) points `lfs.url` |
| at the |
| [Hugging Face mirror](https://huggingface.co/datasets/open-reaction-database/ord-data), |
| so clones and forks fetch LFS objects from HF's CDN instead of GitHub. |
| - **GitHub remains the source of truth.** LFS objects are always written to |
| GitHub (storage there is fine; only download bandwidth was the problem), and |
| the [mirror workflow](.github/workflows/huggingface_mirror.yml) copies them to |
| Hugging Face after every merge to `main`. Hugging Face is purely a read |
| replica — every object is always retrievable from GitHub. |
| - **LFS is scoped to `data/`** (see [`.gitattributes`](.gitattributes)). A new |
| dataset staged at the repository root is an ordinary Git file, so submissions |
| can be pushed from a fork with no LFS configuration; the submission workflow |
| turns the file into an LFS object when it moves it into `data/`. |
| |
| ### For contributors |
| |
| - **Submitting a new dataset:** nothing special is required — stage your file at |
| the repository root and open a PR (see [CONTRIBUTING.md](CONTRIBUTING.md) and |
| the |
| [Submission Workflow](https://docs.open-reaction-database.org/en/latest/submissions.html)). |
| - **Editing a file that already lives under `data/` from a fork:** that file is |
| an LFS object, so point LFS uploads at your own fork once before pushing (you |
| cannot write to the canonical repository's LFS store): |
| |
| ```bash |
| git config lfs.pushurl https://github.com/<your-username>/ord-data.git/info/lfs |
| ``` |
| |
| ### For maintainers (CI) |
| |
| Freshly pushed objects are not on the Hugging Face mirror until the post-merge |
| mirror job runs, so CI and the mirror override the read endpoint back to GitHub |
| at runtime (`git config lfs.url …`): |
| |
| - [`validation.yml`](.github/workflows/validation.yml) pulls only each matrix |
| shard's objects from GitHub, sparsely, instead of the whole dataset in every |
| job. |
| - [`submission.yml`](.github/workflows/submission.yml) reads from GitHub so fork |
| and branch submissions are validated before their bytes reach Hugging Face. |
| - [`huggingface_mirror.yml`](.github/workflows/huggingface_mirror.yml) reads the |
| to-be-mirrored objects from GitHub. |
| |
| ## Contributing |
| |
| Please see the [Submission Workflow](https://docs.open-reaction-database.org/en/latest/submissions.html) documentation. Make sure to review the [license](https://github.com/open-reaction-database/ord-data/blob/main/LICENSE) and [terms of use](https://github.com/open-reaction-database/ord-data/blob/main/CONTRIBUTING.md#terms-of-use). |
| |
| ## Maintainer notes |
| |
| ### Skipping the `Update submission` step |
| |
| The submission workflow's `Update submission` step runs |
| [`scripts/process_dataset.py`](scripts/process_dataset.py) `--update --cleanup` |
| to assign reaction/dataset IDs and timestamps to newly |
| submitted files and rewrite them to the canonical on-disk format. For |
| maintainer PRs that touch dataset files but should *not* be re-processed |
| this way — e.g., format conversions or mass migrations of already-finalized |
| data — apply the `skip-update-submission` label to the PR. Note that the |
| label leaves such a PR with no dataset validation inside the submission |
| workflow: `Validate submission` runs only for fork PRs and for this |
| workflow's own re-runs, so on a labeled maintainer PR both steps sit out. |
| The corpus sweep in [`validation.yml`](.github/workflows/validation.yml) |
| covers the result on merge to `main`. |
|
|
| ### Converting datasets to Parquet |
|
|
| Published datasets are stored as Parquet only, and new submissions are written |
| that way by `process_dataset.py`, so `data/` holds no `.pb.gz` to convert. |
| [`scripts/convert_to_parquet.py`](scripts/convert_to_parquet.py) handles the |
| case where one lands there anyway: it globs `data/*/ord_dataset-*.pb.gz`, |
| converts each 1:1 (carrying the existing `dataset_id`), and refuses to overwrite |
| an existing output whose reaction count disagrees with its input. Converting the |
| corpus is this script's job, not a side effect of editing one dataset, which is |
| why `process_dataset.py` leaves an edited file in whatever format it has. |
|
|
| It needs `ord_schema`, which comes from the `pipeline` dependency group. Because |
| it reads `.pb.gz` content, pull the inputs first. With `data/` Parquet-only the |
| converter has nothing to do and says so — these steps are worth running only |
| once a `.pb.gz` has actually appeared: |
|
|
| ```bash |
| uv sync --only-group pipeline |
| |
| git lfs pull --include="data/*/ord_dataset-*.pb.gz" |
| uv run --no-sync python scripts/convert_to_parquet.py --dry-run # preview |
| uv run --no-sync python scripts/convert_to_parquet.py |
| git rm data/*/*.pb.gz # keep only the Parquet versions |
| ``` |
|
|
| Commit the new `.parquet` files (they become LFS objects), push them (see |
| [Pushing new LFS objects](#pushing-new-lfs-objects)), and open the PR with the |
| `skip-update-submission` label. Validation runs against the full dataset on |
| merge to `main`. |
|
|
| ### Pushing new LFS objects |
|
|
| [`.lfsconfig`](.lfsconfig) routes LFS **reads** to the Hugging Face mirror and |
| deliberately sets no `pushurl`, so a plain push would try to upload new objects |
| to HF — which you cannot write. Point LFS uploads at GitHub for the push, and |
| make sure git can authenticate to `github.com` over **HTTPS** for the LFS API |
| (the LFS endpoint is HTTPS even when your `git` remote is SSH). The simplest |
| auth is the GitHub CLI: |
|
|
| ```bash |
| git config lfs.pushurl https://github.com/open-reaction-database/ord-data.git/info/lfs |
| gh auth setup-git # let git use your gh token for github.com over HTTPS |
| git push -u origin <branch> |
| ``` |
|
|
| Or, as a one-off without persisting any config: |
|
|
| ```bash |
| git -c lfs.pushurl=https://github.com/open-reaction-database/ord-data.git/info/lfs \ |
| -c 'credential.https://github.com.helper=!gh auth git-credential' \ |
| push -u origin <branch> |
| ``` |
|
|
| Reads stay on the mirror; only your uploads go to GitHub. On merge to `main`, |
| `huggingface_mirror.yml` copies the new objects to Hugging Face. |
|
|
| ## License |
|
|
| This repository carries two licenses, because it holds both data and code: |
|
|
| | what | license | file | |
| | --- | --- | --- | |
| | The datasets under `data/` (and the repository metadata describing them) | [CC-BY-SA-4.0](LICENSE) | `LICENSE` | |
| | The code under `scripts/` and `.github/` | [Apache-2.0](LICENSE-CODE) | `LICENSE-CODE` | |
|
|
| If you are using ORD data, CC-BY-SA-4.0 is the license that applies to you; it |
| is what `CITATION.cff` and the [Hugging Face |
| mirror](https://huggingface.co/datasets/open-reaction-database/ord-data) declare. |
|
|
| The code carries a separate license because Creative Commons licenses are not |
| intended for software — [Creative Commons recommends against |
| it](https://creativecommons.org/faq/#can-i-apply-a-creative-commons-license-to-software) |
| — and because the project's other code repositories, including |
| [ord-schema](https://github.com/Open-Reaction-Database/ord-schema), are |
| Apache-2.0. |
|
|
| ## Citation |
|
|
| If you use this dataset, please cite the Open Reaction Database paper: |
|
|
| ```bibtex |
| @article{kearnes2021ord, |
| author = {Kearnes, Steven M. and Maser, Michael R. and Wleklinski, Michael and Kast, Anton and Doyle, Abigail G. and Dreher, Spencer D. and Hawkins, Joel M. and Jensen, Klavs F. and Coley, Connor W.}, |
| title = {The Open Reaction Database}, |
| journal = {Journal of the American Chemical Society}, |
| year = {2021}, |
| volume = {143}, |
| number = {45}, |
| pages = {18820--18826}, |
| doi = {10.1021/jacs.1c09820}, |
| } |
| ``` |
|
|
