File size: 8,544 Bytes
e4c8b1d
 
 
 
 
051d3f3
e4c8b1d
 
051d3f3
e4c8b1d
 
 
 
 
 
0591d06
 
e4c8b1d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d7012da
 
 
051d3f3
 
d7012da
 
 
051d3f3
 
d7012da
 
 
051d3f3
 
d7012da
051d3f3
 
d7012da
 
 
051d3f3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d7012da
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e4c8b1d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d7012da
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
"""Post-merge actions: update manifest.json and dataset card on PR merge."""

import io
import json
import logging
import os
from datetime import datetime, timezone

from huggingface_hub import HfApi, hf_hub_download, snapshot_download
from huggingface_hub.utils import EntryNotFoundError

from dedup import DATASET_REPO_ID, compute_fingerprint, compute_sha256

logger = logging.getLogger(__name__)

api = HfApi()


def _list_data_dirs(api: HfApi) -> list[str]:
    """List top-level directory names under data/ (these become config/split names)."""
    dirs: list[str] = []
    for entry in api.list_repo_tree(
        repo_id=DATASET_REPO_ID,
        repo_type="dataset",
        revision="main",
        path_in_repo="data",
    ):
        if not hasattr(entry, "rfilename"):  # it's a directory
            # entry.path is like "data/global-mmlu-lite"
            name = entry.path.split("/", 1)[-1]
            dirs.append(name)
    return sorted(dirs)


def _build_dataset_card(configs: list[str]) -> str:
    """Build a dataset card README.md with YAML frontmatter for the viewer."""
    yaml_configs = []
    for config in configs:
        yaml_configs.append(f"  - config_name: {config}")
        yaml_configs.append(f"    data_files:")
        yaml_configs.append(f"      - split: train")
        yaml_configs.append(f"        path: data/{config}/**/*.json")

    yaml_block = "\n".join(yaml_configs)

    return f"""---
configs:
{yaml_block}
license: mit
---

# EEE Datastore

Evaluation data for the EEE project.
"""


def update_manifest(api: HfApi, merged_files: list[str]) -> None:
    """Download merged files from main, compute hashes, and update manifest.json."""
    # Load existing manifest
    try:
        manifest_path = hf_hub_download(
            repo_id=DATASET_REPO_ID,
            filename="manifest.json",
            repo_type="dataset",
            revision="main",
        )
        with open(manifest_path, "r") as f:
            manifest = json.load(f)
    except (EntryNotFoundError, Exception):
        manifest = {"files": {}}

    now = datetime.now(timezone.utc).isoformat()

    for file_path in merged_files:
        try:
            local_path = hf_hub_download(
                repo_id=DATASET_REPO_ID,
                filename=file_path,
                repo_type="dataset",
                revision="main",
            )
            with open(local_path, "rb") as f:
                content = f.read()

            sha256 = compute_sha256(content)

            if file_path.endswith(".json"):
                fingerprint = compute_fingerprint(content)
            else:
                fingerprint = sha256

            manifest["files"][file_path] = {
                "sha256": sha256,
                "fingerprint": fingerprint,
                "added_at": now,
            }
            logger.info("Added %s to manifest", file_path)
        except Exception:
            logger.exception("Failed to process %s for manifest", file_path)

    # Upload updated manifest
    manifest_bytes = json.dumps(manifest, indent=2, sort_keys=True).encode()
    api.upload_file(
        path_or_fileobj=io.BytesIO(manifest_bytes),
        path_in_repo="manifest.json",
        repo_id=DATASET_REPO_ID,
        repo_type="dataset",
        commit_message="Update manifest.json after PR merge",
    )
    logger.info("Uploaded updated manifest.json (%d files)", len(manifest["files"]))


def update_dataset_card(api: HfApi) -> None:
    """Regenerate the dataset card README.md with configs for all data/ subdirs."""
    configs = _list_data_dirs(api)
    if not configs:
        logger.warning("No data directories found, skipping dataset card update")
        return

    card_content = _build_dataset_card(configs)
    api.upload_file(
        path_or_fileobj=io.BytesIO(card_content.encode()),
        path_in_repo="README.md",
        repo_id=DATASET_REPO_ID,
        repo_type="dataset",
        commit_message="Update dataset card with configs for viewer",
    )
    logger.info("Updated dataset card with %d configs: %s", len(configs), configs)


def full_rebuild() -> dict:
    """Rebuild manifest.json from scratch and regenerate the dataset card.

    Downloads the entire dataset at once via snapshot_download, then walks
    the local directory to compute hashes/fingerprints for all data files.
    """
    logger.info("Starting full rebuild of manifest + dataset card")

    # Download entire dataset in one shot
    local_dir = snapshot_download(
        repo_id=DATASET_REPO_ID,
        repo_type="dataset",
        revision="main",
    )
    logger.info("Downloaded dataset snapshot to %s", local_dir)

    # Walk local data/ directory to find all data files
    data_root = os.path.join(local_dir, "data")
    now = datetime.now(timezone.utc).isoformat()
    manifest = {"files": {}}

    if not os.path.isdir(data_root):
        logger.warning("No data/ directory found in snapshot")
    else:
        for dirpath, _, filenames in os.walk(data_root):
            for filename in filenames:
                if not (filename.endswith(".json") or filename.endswith(".jsonl")):
                    continue
                local_path = os.path.join(dirpath, filename)
                # Convert to repo-relative path (data/...)
                repo_path = os.path.relpath(local_path, local_dir)
                try:
                    with open(local_path, "rb") as f:
                        content = f.read()

                    sha256 = compute_sha256(content)
                    if filename.endswith(".json"):
                        fingerprint = compute_fingerprint(content)
                    else:
                        fingerprint = sha256

                    manifest["files"][repo_path] = {
                        "sha256": sha256,
                        "fingerprint": fingerprint,
                        "added_at": now,
                    }
                    logger.info("Indexed %s", repo_path)
                except Exception:
                    logger.exception("Failed to index %s", repo_path)

    logger.info("Indexed %d data files total", len(manifest["files"]))

    # Upload fresh manifest
    manifest_bytes = json.dumps(manifest, indent=2, sort_keys=True).encode()
    api.upload_file(
        path_or_fileobj=io.BytesIO(manifest_bytes),
        path_in_repo="manifest.json",
        repo_id=DATASET_REPO_ID,
        repo_type="dataset",
        commit_message="Full rebuild of manifest.json",
    )
    logger.info("Uploaded rebuilt manifest.json (%d files)", len(manifest["files"]))

    # Regenerate dataset card
    update_dataset_card(api)

    return {
        "status": "ok",
        "action": "full_rebuild",
        "files_indexed": len(manifest["files"]),
    }


def handle_merge(pr_num: int) -> dict:
    """Run all post-merge actions for a PR."""
    logger.info("Handling merge for PR #%d", pr_num)

    # Find which data files were added/changed by this PR
    # After merge, everything is on main, so we list all data files
    # and update manifest for any that aren't already tracked
    try:
        manifest_path = hf_hub_download(
            repo_id=DATASET_REPO_ID,
            filename="manifest.json",
            repo_type="dataset",
            revision="main",
        )
        with open(manifest_path, "r") as f:
            manifest = json.load(f)
    except (EntryNotFoundError, Exception):
        manifest = {"files": {}}

    # List all data files on main and find untracked ones
    all_data_files: list[str] = []
    for entry in api.list_repo_tree(
        repo_id=DATASET_REPO_ID,
        repo_type="dataset",
        revision="main",
        recursive=True,
    ):
        if not hasattr(entry, "rfilename"):
            continue
        path = entry.rfilename
        if path.startswith("data/") and (path.endswith(".json") or path.endswith(".jsonl")):
            all_data_files.append(path)

    untracked = [f for f in all_data_files if f not in manifest.get("files", {})]
    logger.info("Found %d untracked data files after merge", len(untracked))

    if untracked:
        update_manifest(api, untracked)

    update_dataset_card(api)

    return {
        "status": "ok",
        "action": "post_merge",
        "pr": pr_num,
        "files_added_to_manifest": len(untracked),
    }


if __name__ == "__main__":
    logging.basicConfig(
        level=logging.INFO,
        format="%(asctime)s %(levelname)s %(name)s: %(message)s",
    )
    result = full_rebuild()
    print(json.dumps(result, indent=2))