Repository navigation
rldb: move MultiDataset and format-neutral pieces out of zarr/ #717
+3,168
−1,780
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,143 @@ | ||
| """Review of annotation change files before `commit.apply_annotations` commits them. | ||
|
|
||
| A change file is a Lance or parquet table with an `op` column ("add" or "delete") and the | ||
| ANNOTATIONS fields. An add row may leave `annotation_id` null, since it is computed from | ||
| the row's content; a delete row needs only `annotation_id`. | ||
| """ | ||
|
|
||
| from collections import defaultdict | ||
| from dataclasses import dataclass, field | ||
| from pathlib import Path | ||
|
|
||
| import lance | ||
| import pyarrow as pa | ||
| import pyarrow.parquet as pq | ||
|
|
||
| from egomimic.rldb.lance.commit import _exists, annotations_uri | ||
| from egomimic.rldb.lance.schema import ANNOTATION_TYPES, ANNOTATIONS, annotation_id | ||
|
|
||
| CONTENT = ("recording_id", "annotation_vendor", "annotation_type", "t_start_ns", "t_end_ns", "text") # fmt: skip | ||
|
|
||
|
|
||
| @dataclass | ||
| class Review: | ||
| adds: pa.Table | ||
| deletes: list[str] | ||
| errors: list[str] = field(default_factory=list) | ||
| unchanged: list[str] = field(default_factory=list) | ||
|
|
||
|
|
||
| def read_changes(path: str | Path) -> pa.Table: | ||
| path = str(path).rstrip("/") | ||
| return ( | ||
| lance.dataset(path).to_table() | ||
| if path.endswith(".lance") | ||
| else pq.read_table(path) | ||
| ) | ||
|
|
||
|
|
||
| def committed(root: Path) -> pa.Table: | ||
| uri = annotations_uri(root) | ||
| if not _exists(uri): | ||
| return ANNOTATIONS.empty_table() | ||
| return lance.dataset(str(uri)).to_table() | ||
|
|
||
|
|
||
| def _task_bounds(root: Path) -> dict[str, list[tuple[int, int]]]: | ||
| cols = ["recording_id", "t_start_ns", "t_end_ns"] | ||
| out = defaultdict(list) | ||
| for t in ( | ||
| lance.dataset(str(Path(root) / "tasks.lance")) | ||
| .to_table(columns=cols) | ||
| .to_pylist() | ||
| ): | ||
| out[t["recording_id"]].append((t["t_start_ns"], t["t_end_ns"])) | ||
| return out | ||
|
|
||
|
|
||
| def _add_errors(r: dict, tasks: dict) -> list[str]: | ||
| errs = [f"{k} is missing" for k in CONTENT if r.get(k) in (None, "")] | ||
| if errs: | ||
| return errs | ||
| if r["annotation_type"] not in ANNOTATION_TYPES: | ||
| errs.append( | ||
| f"annotation_type {r['annotation_type']!r} not in {ANNOTATION_TYPES}" | ||
| ) | ||
| t0, t1 = r["t_start_ns"], r["t_end_ns"] | ||
| if t0 >= t1: | ||
| errs.append(f"t_start_ns {t0} is not before t_end_ns {t1}") | ||
| if r["recording_id"] not in tasks: | ||
| errs.append(f"recording {r['recording_id']} has no tasks") | ||
| elif not any(s <= t0 and t1 <= e for s, e in tasks[r["recording_id"]]): | ||
|
Contributor
Author
There was a problem hiding this comment. medium (correctness): This containment check rejects any annotation that covers a recording's last frame. Annotations are half-open, so such a span ends at ts[-1] + period. But aria_to_lance writes the task's |
||
| errs.append(f"[{t0}, {t1}) is not inside any task of its recording") | ||
| q = r.get("quality_score") | ||
| if q is not None and not 0 <= q <= 5: | ||
| errs.append(f"quality_score {q} is outside 0-5") | ||
| return errs | ||
|
|
||
|
|
||
| def review(root: Path, changes: pa.Table) -> Review: | ||
| """Validate `changes` against the committed tables; nothing is written.""" | ||
| tasks = _task_bounds(root) | ||
| have = set(committed(root).column("annotation_id").to_pylist()) | ||
| adds, deletes, errors, unchanged, seen = [], [], [], [], set() | ||
| for i, r in enumerate(changes.to_pylist()): | ||
| op, given = r.get("op"), r.get("annotation_id") | ||
| if op == "delete": | ||
| if not given: | ||
| errors.append(f"row {i}: delete without annotation_id") | ||
| elif given not in have: | ||
| errors.append(f"row {i}: delete of {given}, which is not committed") | ||
| elif given in seen: | ||
| errors.append(f"row {i}: {given} appears twice") | ||
| else: | ||
| deletes.append(given) | ||
| seen.add(given) | ||
| continue | ||
| if op != "add": | ||
| errors.append(f"row {i}: op {op!r} is not add or delete") | ||
| continue | ||
| errs = _add_errors(r, tasks) | ||
| if errs: | ||
| errors += [f"row {i}: {e}" for e in errs] | ||
| continue | ||
| aid = annotation_id(*(r[k] for k in CONTENT)) | ||
| if given and given != aid: | ||
| errors.append( | ||
| f"row {i}: annotation_id {given} does not match its content ({aid})" | ||
| ) | ||
| elif aid in seen: | ||
| errors.append(f"row {i}: {aid} appears twice") | ||
| elif aid in have: | ||
| unchanged.append(aid) | ||
| else: | ||
| adds.append( | ||
| {**{k: r.get(k) for k in ANNOTATIONS.names}, "annotation_id": aid} | ||
| ) | ||
| seen.add(aid) | ||
| return Review( | ||
| pa.Table.from_pylist(adds, schema=ANNOTATIONS), deletes, errors, unchanged | ||
| ) | ||
|
|
||
|
|
||
| def diff(root: Path, rv: Review) -> str: | ||
| """Per recording: rows added, deleted and already committed, with their texts.""" | ||
| old = {r["annotation_id"]: r for r in committed(root).to_pylist()} | ||
| per = defaultdict(lambda: {"+": [], "-": [], "=": 0}) | ||
| for r in rv.adds.to_pylist(): | ||
| per[r["recording_id"]]["+"].append(r) | ||
| for aid in rv.deletes: | ||
| per[old[aid]["recording_id"]]["-"].append(old[aid]) | ||
| for aid in rv.unchanged: | ||
| per[old[aid]["recording_id"]]["="] += 1 | ||
| lines = [] | ||
| for rid in sorted(per): | ||
| p = per[rid] | ||
| lines.append(f"{rid}: +{len(p['+'])} -{len(p['-'])} ={p['=']}") | ||
| for sign in "-+": | ||
| for r in sorted(p[sign], key=lambda r: r["t_start_ns"]): | ||
| lines.append( | ||
| f" {sign} [{r['t_start_ns']}, {r['t_end_ns']}) " | ||
| f"{r['annotation_vendor']}/{r['annotation_type']}: {r['text']}" | ||
| ) | ||
| return "\n".join(lines) | ||
Oops, something went wrong.
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
nit (naming):
_existsis now used outside commit.py, and the new aria_to_lance_ray.py repeats the same(uri / "_versions").exists()check at lines 83 and 115. Make it public (e.g.commit.table_exists) and use it in both places.