fix(api): validate a new node before deleting the existing one - #4410
Merged
Merged
Conversation
d-v-b
added a commit
to d-v-b/zarr-python
that referenced
this pull request
Sep 25, 2026
Assisted-by: ClaudeCode:claude-opus-5-5
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #4410 +/- ##
==========================================
- Coverage 94.48% 94.47% -0.01%
==========================================
Files 93 93
Lines 13245 13241 -4
==========================================
- Hits 12514 12510 -4
Misses 731 731
🚀 New features to boost your workflow:
|
This was referenced Sep 25, 2026
d-v-b
changed the title
Opening or creating a node with mode="w" or overwrite=True deleted the existing node before the new one was validated, so an invalid argument destroyed the existing data and created nothing. For example, zarr.open(store, mode="w", config=...) on an existing array erased the store and then raised TypeError from open_group. Two deletion points ran too early: - make_store_path(mode="w") erased the path before dispatch in open, open_array, open_group, create and save_array. These now build the store path with mode "a" and create the node with overwrite=True. - overwrite=True deleted the node before its metadata was built in AsyncArray._create_v2/_create_v3, init_array and AsyncGroup.from_store. save_new_metadata now encodes the metadata first, then deletes, then writes. save_group checks its arguments before erasing the path. Assisted-by: ClaudeCode:claude-opus-5-5
Assisted-by: ClaudeCode:claude-opus-5-5
`StorePath.open(mode="w")` no longer deletes the keys under the path. The node created there replaces them through `save_new_metadata(overwrite=True)`, once its metadata is valid, so the API functions pass their mode through unchanged and `_deferred_overwrite_mode` is gone. A read-only store refused with mode "w" names that mode again in its error. `save_group` resolves every array's data type before the group is replaced, and replaces it explicitly with `AsyncGroup.from_store(overwrite=True)`. The docstrings of `open`, `open_group` and `create` state that a store which cannot delete keys raises instead of replacing an existing node; the tests pin that contract over `ZipStore`, seed the overwrite tests with a group that has children, and cover the read-only error and the `save_group` dtype path. Assisted-by: ClaudeCode:claude-fable-5-1 Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
d-v-b
force-pushed
the
fix/validate-before-overwrite
branch
from
September 30, 2026 18:10
87bd5bc to
2531f58
Compare
d-v-b
marked this pull request as ready for review
September 30, 2026 18:35
d-v-b
added a commit
to d-v-b/zarr-python
that referenced
this pull request
Oct 1, 2026
…egular-chunk-grid-4374 Resolves the overlap with zarr-developers#4410: save_new_metadata (main) replaces this branch's encode-then-_prepare_overwrite at the create sites, and save_metadata and save_new_metadata encode through encode_documents so errors name the node. Assisted-by: ClaudeCode:claude-fable-5-1 Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
d-v-b
added a commit
to d-v-b/zarr-python
that referenced
this pull request
Oct 1, 2026
Resolves conflicts with zarr-developers#4410, which replaced _prepare_overwrite with save_new_metadata: keep main's helper and this branch's imports and chunk normalization. Assisted-by: ClaudeCode:claude-fable-5-1 Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
d-v-b
added a commit
to d-v-b/zarr-python
that referenced
this pull request
Oct 1, 2026
…chunk-sizes Resolves the overlap with zarr-developers#4410: save_new_metadata (main) replaces this branch's encode-then-_prepare_overwrite at the create sites, and save_metadata and save_new_metadata encode through encode_documents so errors name the node. Assisted-by: ClaudeCode:claude-fable-5-1 Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
d-v-b
added a commit
that referenced
this pull request
Oct 2, 2026
…by zarr 3.2.x (#4375) * fix(chunk-grids): enforce one zero-length-axis invariant across model, clamps and metadata Invariant: a chunk edge length is always >= 1; a dimension's extent may be 0, in which case the dimension has zero chunks (ceildiv(0, size) == 0). Zero-length-axis bugs have recurred since 2017 (#150, #241, #303, #972, #1977, #2434, #3711, #4305, #4307, #4328) because the layers disagreed on this invariant and every span-derived chunk spelling clamped on its own: - The metadata layer (common.py, metadata/v3.py) required chunk edges >= 1, but the in-memory FixedDimension allowed size == 0 with four special-case branches left over from #2434, so normalization could build a grid the metadata constructor then rejected. FixedDimension now rejects size < 1 and the four `if self.size == 0` branches are gone. VaryingDimension already required edges > 0 and is unchanged. - `chunks=-1`, `chunks=False`, `chunks="auto"` (_guess_regular_chunks, both the typesize == 0 early return and the np.maximum line) and `shards="auto"` each derived "one chunk covering the axis" independently. They now all go through one helper, `_full_span_chunk_size(span) = max(span, 1)`, which is the single definition of that phrase for a possibly zero-length axis. - Zarr format 2 metadata had no chunk >= 1 check, so a legacy `chunks: [0]` document opened fine and read uninitialised memory after a resize. It now raises a clear ValueError at parse time, matching the format 3 grid. - Rectilinear grids had no creation-time spelling for a zero-length axis: normalize_chunks_1d required sum(edges) == span, which no list of positive edges can satisfy for span 0, even though the same state is reachable via resize((0,)) and round-trips through reopen. For span == 0 any non-empty list of positive edges is now accepted verbatim, producing the same VaryingDimension(edges, extent=0) that resize produces; the strict sum check is kept for span > 0. Tests: the per-spelling regression test from #4328 is replaced by one matrix over {-1, False, "auto", 1, (1,...), [[2, 2]]} x {(0,), (0, 4), (4, 0), (0, 0), ()} x {v2, v3} x {no shards, shards="auto" with and without a byte budget, explicit shards}, with separate small tests for each error case. Tests that constructed FixedDimension(size=0) now assert it raises, and a zero-extent test covers the behaviour the old special cases were guarding. Assisted-by: ClaudeCode:claude-fable-5-1 * fix(metadata): read a legacy v2 zero chunk edge on an empty axis as 1 zarr-python 2.18.7 writes `chunks: [0]` for `zarr.zeros((0,), chunks=False)` and for `chunks=(0,)`, so stores with that document exist. Rejecting them at open would turn a previously-readable array into an error; leaving the 0 in place read uninitialised memory after a resize. Normalize the edge to 1 with a ZarrUserWarning instead — the same grid every other "one chunk spans the axis" spelling produces — and keep rejecting a zero edge on an axis that has data. Assisted-by: ClaudeCode:claude-fable-5-1 * chore: rename changelog fragment to the PR number Assisted-by: ClaudeCode:claude-fable-5-1 * docs: state what 2.x actually did with a zero chunk edge Measured against zarr 2.18.7: `zeros((0,), chunks=False)`, `chunks=-1` and `chunks=(0,)` all write `chunks: [0]`, after which nchunks, read, write, append, resize and reopen-then-read every raise ZeroDivisionError. There was never a working behaviour to preserve; normalizing the edge to 1 makes such arrays usable for the first time. Say so in the comment and fragment instead of claiming the stores were previously readable. Assisted-by: ClaudeCode:claude-fable-5-1 * docs: qualify zero-chunk compatibility history Assisted-by: Codex:GPT-6 * fix: read mixed regular/rectilinear chunk grids written by zarr 3.2.x zarr 3.2.0 and 3.2.1 classified mixed chunk specs like (2, (5, 10, 5)) as regular and stored them as a "regular" grid whose chunk_shape contains an edge list, while laying the chunks out as a rectilinear grid. Newer versions accepted that metadata and failed later with an unrelated TypeError. RegularChunkGridMetadata now rejects edge lists. When reading stored metadata, a "regular" grid with edge lists is read as the rectilinear grid it describes, with a warning explaining how to re-save it; if rectilinear chunks are disabled, the error says what happened and how to enable them. Closes #4374 Assisted-by: ClaudeCode:claude-opus-5 * docs: add changelog fragment for #4375 Assisted-by: ClaudeCode:claude-opus-5 * fix: accept numpy integers in regular chunk grids; normalize tuples in the 3.2.x shim Review follow-ups for #4375: - RegularChunkGridMetadata now accepts numpy integer scalars and stores Python ints, mirroring parse_shapelike. The strict integer check had reported np.int64(2) as if it were a list of chunk edges. - The mixed "regular" grid reader converts tuple entries to lists before delegating to RectilinearChunkGridMetadata.from_dict, so metadata dicts built in Python are handled the same as parsed JSON. - The error and warning share one message prefix. - Tests cover numpy ints, run-length encoded edges, and tuple input, and the test section header says the reader is broader than the 3.2.x bug. Assisted-by: ClaudeCode:claude-fable-5-1 * refactor: branch on integer vs sequence when parsing rectilinear chunk shapes The chunk_shapes parsers checked exact types: `from_dict` accepted a dimension only as `int` or `list`, `expand_rle` accepted an RLE pair only as a `list`, and the reader for 3.2.x mixed grids special-cased `tuple` to convert it to a list before handing it to `from_dict`. A metadata dict built in Python holds tuples where parsed JSON holds lists, and a numpy array is as good as either, so these checks rejected valid input. One structural predicate, `declares_chunk_edges`, now answers "is this a sequence of edges rather than a single integer size" for all of them: any non-integer iterable except `str`/`bytes`. It is a `TypeGuard`, not a `TypeIs`, because it is False for strings, which are iterable. The four sites that make that decision use it: `parse_chunk_grid`'s detection of a mixed grid, the mixed-grid reader, `RectilinearChunkGridMetadata.from_dict`, and `expand_rle`. The shim no longer needs its tuple special case. Integers are `int | np.integer` throughout, matching `_parse_chunk_shape`, and every parser stores Python ints: `_validate_chunk_shapes` now coerces, so constructing a rectilinear grid from numpy values works the way it already did for a regular grid. Assisted-by: ClaudeCode:claude-opus-5 * docs: note the rectilinear parsing change in the changelog fragment Assisted-by: ClaudeCode:claude-opus-5 * refactor: validate each chunk grid kind once, in one place Parsing a regular chunk shape repeated its work at two levels: - `_parse_chunk_shape` type-checked and coerced every dimension, then handed the result to `_validate_chunk_shapes`, which re-ran the same isinstance test and coercion and added only the `>= 1` check. Its edge-list branch was unreachable from this caller, which is why a `cast` was needed on the way out. - `RegularChunkGridMetadata.from_dict` parsed the chunk shape and passed it to the constructor, whose `__post_init__` parsed it again. Together that was four passes over the dimensions for `from_dict`. It is now one: `_parse_chunk_shape` checks the range itself and no longer calls the rectilinear validator, and `from_dict` hands the dimensions to the constructor unparsed. Beyond the redundancy, the shared validator was the coupling that let a rectilinear chunk shape be stored as a regular grid (#4374), so the two grid kinds now validate separately. `RectilinearChunkGridMetadata.from_dict` also had its own `>= 1` check for bare-int dimensions, duplicating `_validate_chunk_shapes`, which `__post_init__` runs over the result anyway. It now only puts the JSON into shape (integer vs sequence, RLE expansion), and a bad bare int is reported by the validator, which names the dimension. `expand_rle` keeps its own checks because it is called directly. Assisted-by: ClaudeCode:claude-opus-5 * fix(metadata): read a legacy zero chunk size on an empty axis in Zarr format 3 The compatibility policy for a stored chunk size of 0 on a zero-length axis covered Zarr format 2 only, so arrays written by zarr-python 3.0 and 3.1 with `chunk_shape: [0]` — or `[false]`, which 3.0 wrote for `chunks=False` — still could not be opened at all. `ArrayV3Metadata` now applies the same policy to a regular chunk grid: a stored chunk size of 0 on a zero-length axis is read as 1 with a `ZarrUserWarning`, and a zero chunk size on a positive-length axis is left for the chunk grid parser to reject. It runs in `__init__` rather than in the grid parser because the policy needs the array shape, which chunk grid metadata does not carry. Both warnings now say how to store a corrected chunk size — open the array writable and call `array.update_attributes({})`, which rewrites the whole document from the parsed metadata — from one shared constant. The Zarr format 2 warning also named only zarr-python 2.x; measured against real installs, every 3.x release before 3.4 wrote a zero chunk size for an empty array too (3.3.0 for `chunks=-1` and `chunks=False`). Tested against stores written by zarr 3.0.10, 3.1.6 and 3.3.0: they open, append without losing data, and re-save to a chunk size that reopens without a warning. Assisted-by: ClaudeCode:claude-opus-5 * refactor(metadata): check stored chunk shapes against the array shape in one routine The legacy zero-chunk policy was written out twice: inline in `ArrayV2Metadata.__init__`, and again in a Zarr format 3 helper. The V2 copy also zipped with `strict=False` and re-appended any trailing chunk entries only so that a separate length check, `parse_metadata`, could report a dimensionality mismatch after construction. `parse_stored_chunk_shape` in `zarr.core.metadata.common` is now the one place a stored chunk shape is checked against its array's shape, for both formats: one entry per axis, every integer chunk size at least 1, and a size of 0 (or JSON `false`) on a zero-length axis read as 1 with a warning that names the writer and how to re-save. Non-integer entries, such as edge lists, pass through for the caller's own parser. `ArrayV2Metadata.__init__` calls it directly and `parse_metadata` is gone. The Zarr format 3 adapter only locates a regular grid's `chunk_shape` in the stored document and hands it over; it still runs in `ArrayV3Metadata.__init__` because chunk grid metadata has no array shape. The Zarr format 3 import changes that existed only for the old helper are reverted. Tests for the policy now target the routine: one table of valid and legacy inputs, and one test per rejection (dimension mismatch, zero on a non-empty axis, negative). They replace metadata-level tests in test_v2.py and test_v3.py that only re-tested the same rules; the end-to-end tests still cover both formats' wiring against stored arrays. Assisted-by: ClaudeCode:claude-opus-5 * refactor(metadata): scope the stored chunk shape check to regular chunk grids `parse_stored_chunk_shape` passed non-integer entries through "for the caller's own parser", which made a regular-grid policy look like a general chunk shape routine and let it decide what a 0-length chunk means for grids it does not own. A rectilinear grid, or any other grid, is free to define its own semantics for 0-length chunks. It is now `parse_stored_regular_chunk_shape`, typed `Sequence[int]`, with no pass-through, and its docstring says it applies to Zarr format 2 `chunks` and Zarr format 3 `regular` grids only. The Zarr format 3 caller hands it a chunk shape only when the grid is named `regular` and every entry is an integer (`_is_regular_chunk_shape`); anything else is not a regular chunk shape and goes to the chunk grid parser untouched. Assisted-by: ClaudeCode:claude-opus-5 * refactor(metadata): read both legacy regular chunk grid forms in one place #4334 read a zero chunk size on an empty axis in `ArrayV3Metadata.__init__` and #4375 read a 3.2.x mixed grid inside `parse_chunk_grid`, each with its own predicate, warning text and re-save advice. Both are compatibility readings of a stored `regular` grid, so `_read_stored_regular_chunk_grid` now dispatches to both, `parse_chunk_grid` accepts only what the spec allows, and both warnings use `RESAVE_METADATA_HINT`. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(metadata): read a stored zero chunk size on a grown axis as one spanning chunk A stored chunk size of 0 was tolerated only on a zero-length axis and rejected otherwise, because the metadata supposedly could not say how the stored chunks were laid out. But a chunk size of 0 gives a grid of zero chunks, so no release could store a chunk under it, and the writers of that metadata let the axis grow: 3.4.0 appends to a Zarr format 2 array created empty by 3.3.0 (shape grows, no chunk written), and 3.1.6 records a Zarr format 3 resize and the resize half of a failed append. Measured with real installs. Those arrays open in 3.4.0, attributes included; the rejection would have made them unopenable. `parse_stored_regular_chunk_shape` now reads a stored 0 (or JSON `false`) on any axis as one chunk spanning it, `max(extent, 1)`, which is what the `-1`/`False` spec that wrote it meant. On a grown axis the warning also says that data written to it was not saved. Negative sizes are still rejected. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * test: a stateful test of one array's create/append/resize/write life The only state machine that touched arrays compared zarr on one store with zarr on a MemoryStore, so a chunk grid bug showed up identically on both sides; it had no append rule, covered Zarr format 3 only, and kept every empty axis at 0 when resizing, which is where the zero-length bugs live. `ArrayLifecycle` checks one array against a NumPy model across both formats, every chunk spelling (-1, False, "auto", ints, sharded, rectilinear) and the stored chunk size of 0 that releases before 3.4 wrote, including on an axis those releases grew. Rules append, resize (growing and shrinking to and from 0), write and re-save the metadata; the invariant reopens the array and compares shape, values and whether the legacy warning is due. Deliberately breaking the grown-axis policy, the legacy warning, or append on an empty axis each fails it. `resize` keeps partly retained chunks whole, so cells cut off by a shrink can come back with their old values when the axis grows (as in 2.x); the model marks such cells unknown until written instead of encoding that. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(metadata): stop the zero chunk size warning from naming writers The warning for a stored chunk size of 0 said which zarr-python releases wrote it. The check runs on every metadata construction, so metadata built in code, such as VirtualiZarr's kerchunk writer passing an empty array's shape as `chunks`, was told it came from zarr-python 2.x. The warning now says what the chunk size is read as and, on an axis of positive length, that the axis holds only the fill value. `legacy_writers` is gone, and the docstrings no longer narrate release history; the changelog keeps it. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(metadata): stop the mixed chunk grid messages from naming releases The warning and error for a stored `regular` grid that lists chunk edges said which zarr releases wrote it. They now say what is wrong with the metadata, that it lists chunk edges only a rectilinear grid can declare, and how it is read. The reader's docstrings no longer narrate release history either; the changelog fragment keeps it. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * chore(metadata): drop an orphaned comment in ArrayV2Metadata.__init__ The comment introduced a consistency check (`parse_metadata`) that this branch removed; the check now lives in `parse_stored_regular_chunk_shape`. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(chunk-grids): size a full-span shard as a multiple of the inner chunk One definition of "one chunk spanning the axis", full_span_chunk_size(span, unit), is the smallest positive multiple of unit covering span. shards=-1 and shards=False now use the inner chunk size as unit, so a zero-length axis, or one whose length is not a multiple of the inner chunk, gets a valid shard instead of a divisibility error. The zero-length array test drops the explicit-size spellings and keeps the spellings that derive a chunk from the span. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * refactor(metadata): read invalid stored chunk sizes in one module of document upgrades The metadata constructors are strict again: a chunk edge length is an int of at least 1, so metadata built in code with 0 or True raises, without a warning. Stored documents are read leniently only in zarr.core.metadata.upgrades, whose upgrades map a stored array document to a valid one before the constructors run. ArrayV2Metadata.from_dict and ArrayV3Metadata.from_dict apply them, so opening an array and parsing consolidated metadata take the same path. The first upgrade reads a regular chunk size of 0 or JSON false as one chunk spanning the axis, a multiple of the inner chunk size when the array is sharded (this opens sharded stores with an outer chunk size of 0), and JSON true, which real releases wrote for chunks=(True,), as 1. Its one warning says what is invalid, how it is read, and how to re-save, including zarr.consolidate_metadata. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * docs: rewrite the 4334 changelog fragment and trim restating docstrings The fragment now covers the full-span shard rule, strict constructors, and the stored-document upgrade (including JSON true) in two short paragraphs. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * test: model exactly what the store holds in the array lifecycle state machine The model now tracks cells beyond the array's shape: a shrinking resize deletes exactly the chunks outside the new grid, kept chunks keep their out-of-bounds cells, and a write covering every in-bounds cell of an unsharded chunk resets its out-of-bounds cells to the fill value. A resize that never deletes chunks now fails the test. Sharding is its own chunk spelling, a stored chunk size of 0 applies to sharded arrays too (in the outer grid), re-saving metadata only runs when the store warns, and the test takes its settings from the repository's hypothesis profiles under the slow_hypothesis marker, like the other stateful tests. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * refactor(metadata): warn about an upgraded document only once it validates An upgrade now returns the upgraded document and how it read it, and `from_dict` warns with those readings after the metadata constructor accepts the upgraded document. A stored document that is still invalid after an upgrade raises its own error instead of first warning how it was read. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(metadata): read a regular grid listing chunk edges as a document upgrade A `regular` chunk grid whose `chunk_shape` holds flat lists of integer edge lengths, e.g. `[2, [5, 10, 5]]`, is now read by one more entry in `V3_ARRAY_UPGRADES` instead of by a reader in `ArrayV3Metadata.__init__`. It applies on every path that parses a stored document, including consolidated metadata, and does not require `array.rectilinear_chunks`. Only what such documents hold is upgraded (integer sizes and flat integer edge lists); run-length encoded or non-integer entries are left for the regular grid to reject. The warning quotes a bounded prefix of the chunk shape and is emitted only after the upgraded document validates. The rectilinear flag moves off the chunk grid metadata constructor to the stored-document boundary: `ArrayV3Metadata.from_dict` checks the document as stored (before upgrades) and `ArrayV3Metadata.to_buffer_dict` checks the document about to be stored. Reading or creating a genuine rectilinear array still requires the flag, as before; re-saving an upgraded array stores a rectilinear grid and so requires it too, which the warning says. Tests: the upgrade's table test and one test per rejected form, using a document copied verbatim from a store zarr 3.2.1 wrote. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * refactor(metadata): drop numpy and tuple widening from chunk grid metadata The chunk grid metadata classes take typed Python values again: a regular chunk shape of `int`s, rectilinear dimensions of `int` or `list` in stored documents. `declares_chunk_edges` and the numpy and tuple handling in `expand_rle`, `_validate_chunk_shapes` and `RectilinearChunkGridMetadata.from_dict` are removed with their tests; numpy input belongs to the `chunks=` normalizer. The regular grid still rejects edge lists (gh-4374), now naming the offending type instead of echoing the value. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * docs: describe reading mixed regular grids without the rectilinear flag Rewrite the 4375 changelog fragment for the upgrade and the moved flag gate, and drop the numpy/tuple parsing claims that no longer hold. Trim the history from the regular chunk shape docstring. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * refactor(metadata): warn about an upgraded document only once it validates An upgrade now returns the upgraded document and how it read it, and `from_dict` warns with those readings after the metadata constructor accepts the upgraded document. A stored document that is still invalid after an upgrade raises its own error instead of first warning how it was read. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * refactor(metadata): one per-axis rule for stored chunk sizes, one rule for edges Stored regular chunk shapes, in both formats and in a sharding codec's inner chunk shape, are read entry by entry by one rule in upgrades.py: an int >= 1 is kept, `true` is read as 1 (zarr 3.0.10 also wrote it as the inner chunk size of a shard, which the sharding codec used to read silently), and 0 or `false` as one chunk spanning the axis. Integral floats are not read: no release wrote a regular grid with them. A document warns once, naming the array where the caller knows its path. Metadata constructors check every chunk edge length with one function, `parse_chunk_edge`, which tells a wrong type (TypeError) from a wrong value (ValueError); this covers bare sizes, rectilinear edges, RLE sizes and the sharding codec's inner chunk shape. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * refactor(chunk-grids): start chunk guessing from the one full-span rule `_guess_regular_chunks` clamped zero-length axes with `np.maximum`, restating `full_span_chunk_size` with unit 1. Fold the zero-span normalizer tests into the `normalize_chunks_1d` tables. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * test: run the array lifecycle state machine in the slow Hypothesis job Add it to `just hypothesis`. Re-saving metadata is always enabled and must leave valid metadata unchanged; appends favour an axis stored with chunk size 0, which raises how often a legacy axis grows before it is re-saved. A fixed example pins that a write to a shard kept by a shrink leaves its cells beyond the shape alone. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(metadata): read a regular grid mixing sizes and edge lists per axis A stored regular chunk shape is read by the one per-axis rule of upgrades.py; a flat list is one more case of it, kept as that axis's chunk edges. A shape mixing chunk sizes and edge lists is read as the rectilinear grid it describes, so `[true, [5, 10, 5]]`, which zarr 3.2.x wrote and read, reads as `[1, [5, 10, 5]]`. A shape of only lists was never stored in a regular grid and is rejected. The rectilinear grid's constructor checks each edge with the one edge rule, so a float or bool in an edge list is reported as itself. The warning says that re-saving stores the rectilinear grid and so needs the flag, before the re-save steps it lists. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(array): check that metadata can be stored before touching the store `check_storable` is the one store-write gate for the rectilinear chunks flag. `ArrayV3Metadata.to_buffer_dict` calls it, and so does `GroupMetadata.to_buffer_dict` for each array in the group's consolidated metadata, so `consolidate_metadata` can no longer store a rectilinear grid with the flag off and leave the group unreadable without it. Every metadata write serializes before it writes. `resize` used to delete the chunks outside the new shape before storing the new metadata, so metadata that could not be stored lost data; it now stores the metadata first, for all metadata, and deletes afterwards. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(array): store upgraded metadata before writing the first chunk An array read from a stored document that the upgrades had to correct (a chunk size of 0, `false` or `true`) wrote chunks under the corrected layout while the store kept the old document, so zarr 3.2.1-3.4.0 then read only the fill value, and tensorstore and zarrs could not open it. `from_dict` now marks the metadata it read from an upgraded document (`_stored_document_upgraded`, a field outside the document and equality), and `AsyncArray._set_selection`, which every chunk write goes through (`setitem` now included), stores that metadata first and keeps the stored copy. Concurrent first writes store the same document. The array lifecycle state machine now ends the legacy state on a write and checks that no chunk is stored under a document that is still upgraded on read; `resave_metadata` runs only while the stored document is invalid, and re-saving valid metadata is checked once, at creation. The changelog also says which releases stored a chunk size of 0 on an axis of positive length. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * refactor(metadata): name an array one way in upgrade warnings Warnings about an upgraded document name the array by its store path, `str(StorePath(store, path))`, on every path that reads one: opening an array, `AsyncArray.from_dict`, a group read with or without consolidated metadata, and consolidated members, which were named relative to the group. `GroupMetadata.from_dict` and `ConsolidatedMetadata.from_dict` take the group's path for that. The consolidated test now also opens the group with `use_consolidated=False` and checks each array's name, so dropping the path on either read path fails a test. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * refactor(metadata): one rule for bare sizes and edge lists; one noun in messages - `_validate_chunk_shapes` decides "edge list or bare size" with one `match`: a list or tuple is an edge list, anything else is a bare size checked by `parse_chunk_edge`, so a string dimension such as `"10"` is rejected as not an int instead of being iterated per character. Stored rectilinear `from_dict` makes the same JSON split and passes the dimension to `expand_rle`, so an invalid edge or RLE count inside a list names its dimension. - Messages say "dimension" throughout, lowercase after the colon ("Dimension 0: chunk edge length must be >= 1, got 0"); the upgrade warnings read "0 in dimension 0 as one chunk spanning the dimension". - The upgrades test for JSON arrays with `list` only (a stored document holds no tuples), and `_read_chunk_size` says what `span=None` means. - `ArrayV2Metadata.from_dict` copies the document once. - The rejected stored chunk shapes get one test per error case. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix: encode metadata before deleting store content `Array.resize`, `AsyncGroup.delitem` and `create_hierarchy(overwrite=True)` now encode every document they will store (so the rectilinear gate runs) before deleting anything, then store the encoded documents. A group member deletion or hierarchy overwrite that would store a rectilinear grid with the flag off no longer deletes data first. `_resize` goes back to deleting before storing, which restores the behaviour of stores that cannot delete (a ZipStore resize fails before writing duplicate metadata). The gate error raised from `GroupMetadata.to_buffer_dict` names the consolidated member. Tests: one table of operations that must leave the store untouched without the flag (resize, chunk write, member deletion, hierarchy overwrite), the consolidation test for a member in a subgroup, a write with the flag that stores the rectilinear document, an abbreviated long chunk shape, and a ZipStore resize. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * test: reject a run-length entry that is not a pair Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(array): store the upgrade of the current stored document before writing A handle read from an upgraded document upserts the upgrade of what the store holds on its first non-empty chunk write: a stale handle no longer rolls back newer metadata, and an empty write stores nothing. consolidate_metadata does the same for each upgraded member before writing the consolidated document. Both go through two internal primitives in zarr.core.metadata.io: diff_documents compares a node's stored documents with those its metadata would store, value by value, and upsert_metadata encodes first, then stores only the documents that differ and returns the changes. The upgraded-document flag is a ClassVar set on the instance by one helper, mark_upgraded, which also warns; the upgrades are keyed by Zarr format; the Metadata.to_dict and V2 from_dict changes are reverted. _append goes through AsyncArray.setitem and _setitem is removed. A chunk shape that is not a list or tuple is rejected as a whole, and every expand_rle error names its dimension. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * test: prefer zero-length axes for stored chunk size 0 in the lifecycle machine An empty write stores no metadata, so it no longer ends the legacy state. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * refactor: the caller that knows the node names it in gate errors `check_storable` and `_check_rectilinear_chunks_enabled` take no subject. `encode_documents(store_path, metadata)` in `metadata/io.py` encodes the documents an operation stores and, on failure, adds a note naming the node at `store_path` in the full-path form the upgrade warnings use; every encode-before-write site goes through it (`save_metadata`, `upsert_metadata`, `_resize`, `Group.delitem`, group attribute updates, `_encode_nodes` for `create_hierarchy`/`create_nodes`). `ArrayV3Metadata.from_dict` names the array at `path` when the flag refuses a rectilinear document, which is where `consolidate_metadata` and a first write to an upgraded 3.2.x mixed grid are refused without the flag. The upsert gate test now uses the real gate instead of a monkeypatched `to_buffer_dict`, and consolidating a 3.2.x mixed grid is checked both ways: refused with the store byte-identical without the flag, stored as rectilinear (member and consolidated document) with it. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(group): delete a consolidated member in place, after encoding without it `Group.delitem` encodes the group documents from a copy without the member (so the gate runs with the store and the group untouched on failure), deletes the member, then pops it from the shared consolidated metadata in place, as main does, so a parent or subgroup handle sharing that consolidated metadata sees the deletion, then stores the documents. Tests: the consolidated delitem test runs for both Zarr formats and reopens from the store (V2 `.zmetadata` is rewritten); a subgroup handle read from its parent's consolidated metadata sees a deletion made through it; a deletion refused without the flag leaves the group listing the member. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(metadata): keep accepting the chunk sizes zarr 3.4.0 accepted in metadata constructors (patch release) Nothing in a patch release may reject a chunk size that zarr 3.4.0 accepted. The one chunk edge rule (`parse_chunk_edge`, which also reads RLE repeat counts) now reads any integral number as the `int` it equals: an `int`, a `bool`, a NumPy integer or an integral float, so rectilinear grids written with float edges (`[[4.0, 2]]`) read again and the grid constructors accept NumPy integers; a fractional float is still rejected. A regular chunk shape may again be any iterable. `ArrayV2Metadata(chunks=...)` and `ShardingCodec(chunk_shape=...)` read their chunk shape with `parse_shapelike`, as in 3.4.0: a scalar, NumPy integers, bools and 0 are accepted, and a 0 is written back as given (the kerchunk writer in VirtualiZarr relies on `chunks=(0,)` for empty inlined arrays). Stored documents with a chunk size of 0 are still read by the upgrades. The strict rule returns in the next minor release. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(array): leave a valid stored document as written on a stale handle's first write A handle read from an upgraded document re-reads the stored document before its first chunk write. If that document no longer needs an upgrade, store nothing: re-encoding it could differ from how another implementation wrote it, and nothing about it needs fixing. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * docs: describe the 4334 changes to metadata constructors for a patch release Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * test: expect the patch release's wording for non-integral chunk edges (patch release) The chunk edge rule reads any integral number (loop/4334), so a float or bool edge in a stored mixed regular grid is read as the int it equals; a fractional or string edge is still rejected, with "must be an integer". Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(metadata): read float edges only in stored rectilinear documents The patch release widened the one chunk edge rule to accept integral floats so that stored rectilinear grids written with float edges (`[[4.0, 2]]`, as zarr 3.2 wrote for float edges) keep opening. That also made a stored regular chunk shape `[10.0]` open, which zarr 3.4.0 rejected and the next minor release would reject again. The constructor rule now accepts integers of any integer type (`int`, `bool`, NumPy integers) and rejects floats. A new document upgrade reads an integral JSON float of at least 1 as an `int` where zarr 3.2 stored one: the explicit edges and run-length encoded sizes of a rectilinear chunk grid, with the standard warning. Bare sizes, repeat counts, regular chunk shapes and inner chunk shapes stay for the constructors to reject. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * test: read float edges in a regular grid that lists chunk edge lengths zarr 3.2 stored `[2, [5.0, 10.0, 5.0]]` for float edges given to a regular grid. The mixed-grid upgrade reads it as a rectilinear grid, and the float edge upgrade that follows reads its edges as integers, both in one warning. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * test(codecs): pickle a ShardingCodec with an inner chunk size of 0 Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(group): build every node before create_hierarchy deletes or stores anything A node whose metadata no array or group can be built from now fails with the store untouched, even with overwrite=True. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(metadata): warn only where the user must act; guard and refresh upgraded copies - Readings that give what zarr 3.4.0 read (0/false on an empty axis, JSON true, float rectilinear edges) are silent; the metadata is still marked upgraded, so the first write re-saves it. - JSON true is read as 1 in stored rectilinear edges and RLE sizes and in nested sharding codecs' inner chunk shapes. - A write through a handle whose stored document now lays out chunks differently raises and stores nothing; with no stored document it writes as before. - Storing group metadata refreshes each upgraded consolidated member from its own stored document (the one place: save_metadata); consolidate_metadata no longer needs its own pass. - Metadata built in code with chunk size 0 is read through the upgrades, so an array can be built from it, as in 3.4.0. - Chunk edges in metadata follow the 3.4.0 integer rule (int; bool read as its value); a string or mapping is rejected as a chunk shape as a whole. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * test: expect the lifecycle machine's upgrade warning only on non-empty axes Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(group): encode a group's consolidated metadata before storing member upgrades After the merge, storing group metadata refreshed each upgraded consolidated member and stored its upgrade before encoding the group, and `Group.delitem` (which encodes before deleting) skipped the refresh. `encode_node` refreshes the consolidated members (reading only) and encodes the group, so the rectilinear gate runs for every member before anything is stored; `store_node` then stores the group's documents and the members' upgrades. `save_metadata` and `Group.delitem` both use the pair, so a group write refused without the flag stores no member upgrade either. Tests: group attribute writes join the store-untouched table; a refused group write stores no other member's upgrade; with the flag, deleting a member or setting a group attribute stores the rectilinear grid in the member and consolidated documents. Error-message expectations follow the patch release's integer rule wording, and the mixed-grid rows expect the JSON `true` and float edge readings to be silent. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(metadata): leave a sharded chunk size of 0 unread when the inner shape is invalid A stored outer chunk size of 0 of a sharded array is read in multiples of the inner chunk size. When that inner size is itself 0 or `false`, the unit is unknown: the upgrade now leaves the 0 for the constructor, which rejects it with the same `ValueError` zarr 3.4.0 raised, instead of a ZeroDivisionError. `_read_codec` now matches the sharding codec like `_inner_chunk_shape` does. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * perf(metadata): read each upgraded member once, concurrently, and adopt it A group handle kept the consolidated copies of upgraded members flagged, so every later group write re-read every such member, one after another, and read each twice on the first write (once to parse, once to diff). - `save_metadata` returns the metadata it stored; `AsyncGroup._save_metadata` (attrs, `update_attributes`, `delitem`, `consolidate_metadata`) and `Group.update_attributes_async` adopt it, so later writes read no member. - The refresh visits only nested groups and flagged members, and reads them with one `asyncio.gather`. The documents it reads are the ones the upsert diffs against (`read_documents` + `upsert_metadata(..., stored)`), and the array write path does the same. - `parse_stored_array` (documents -> silently marked metadata) replaces `read_stored_array`'s `(metadata, bool)` tuple, and is the one "read, mark, don't warn" path, also for code-built metadata. - A flagged member whose own document is gone or cannot be read (invalid, replaced by a group) keeps its consolidated copy instead of failing every group write. - `parse_array_metadata` of a metadata object reads it as a stored document only when `ArrayV2Metadata.chunks` holds a 0, the one chunk size that constructors still accept and the upgrades change. This removes the per-construction `to_dict` + upgrade (AsyncArray() back to 3.4.0 speed) and building arrays from codec configurations holding NumPy scalars works again, as in 3.4.0. - `create_hierarchy` documents that it stores a group's consolidated metadata as given. - Tests: second group write reads no member; nested member refresh; member without a readable document; NumPy-scalar codec configuration; stale handle whose inner chunk shape alone changed (kills `_chunk_layout` returning only the outer grid). Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(group): store the consolidated metadata as it is after a deletion `delitem` on a consolidated group stored the group metadata it had encoded before deleting, so concurrent deletions through one handle each stored a member list that still held the others' members. The up-front encode now only validates (metadata that cannot be stored still fails with the store untouched); after the deletion the handle's current metadata is stored and adopted, as zarr 3.4.0 stored it. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * refactor: name a node in errors as "Array 'path': what happened" The rectilinear flag's read note, the stale-handle write error and the encode note now share the shape the stored-document warnings use: the node, its path, then what was (not) done. The consolidated-member note keeps its location qualifier ("in the consolidated metadata"), followed by the group's note. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(group): adopt refreshed consolidated members in place A group write replaced the handle's metadata with a copy holding the consolidated members it read again, so the handle no longer shared its consolidated dicts with subgroup handles taken earlier: an array deleted through such a subgroup was stored again by the parent's next write, and reappeared on reopen. `_refresh_consolidated` now returns, along with the copy it encodes, the members it replaced (with the consolidated dict that holds each), and `save_metadata` writes them back into those dicts once the store succeeds, unless the member was deleted or replaced meanwhile. `save_metadata` returns nothing again, and the group callers keep their metadata objects. A failed store leaves the members flagged, so a retry reads them again. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(metadata): raise the rectilinear flag error when refreshing a consolidated member A group write reads each upgraded consolidated member again from its own document, and kept the consolidated copy when that document could not be read. That also swallowed the rectilinear chunks flag error, so a member whose document now declares a rectilinear grid had its stale copy stored as valid. The flag error is now `RectilinearChunksDisabledError`, a `ValueError`, and the refresh lets it propagate: the group write raises and stores nothing. Unreadable documents keep their consolidated copy as before. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(group): say nothing was stored when a consolidated member needs the flag Storing a group's consolidated metadata reads each upgraded member again from its own document; one read as a rectilinear chunk grid now raises the flag error there (naming the array) instead of being kept stale. That error now also says the group stored nothing, as the encoding error does. `zarr.consolidate_metadata` without the flag reports the member this way. The refresh test gains a mixed regular grid row. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * perf(group): read each upgraded member once when deleting a member `delitem` encodes the group metadata without the member before deleting it, which reads each upgraded consolidated member again, then stored the metadata through `save_metadata`, which read them all a second time. It now adopts the members the first encoding read, after the deletion and in place, and stores the metadata as it then is with those members' upgrades (`store_node`), so the first deletion reads each member once. The handle's consolidated metadata is no longer replaced by writes, so the deletion pops from it directly. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(group): store a group's documents only; keep upgraded members as stored Group writes (attributes, `update_attributes`, deleting a member, `consolidate_metadata`, `create_hierarchy`) read each upgraded member of the consolidated metadata again from its own document and stored its upgrade. Those member writes raced concurrent deletions: two concurrent `delitem`s on a consolidated group recreated a deleted array's metadata document in most runs, which zarr 3.4.0 never did. A group write now stores only the group's own documents and reads none. Array metadata read from a document that had to be upgraded keeps that document (`_stored_document`, set by `mark_upgraded`, replacing the `_stored_document_upgraded` flag), and consolidated metadata stores such a member as it was stored. Every reader of the consolidated metadata then reads the member as upgraded again, and the array's own first chunk write stores the upgrade of its current document (or refuses a changed chunk grid) as before: the only write that upgrades a member document. Removed: the refresh and adoption of consolidated members in `save_metadata` (`_refresh_consolidated`, `_refresh_array`, `_Refreshed`), and `Group.update_attributes_async`'s detour through `save_metadata`. Tests pinning the refresh are replaced by tests that a group write reads nothing and writes only the group's documents, stores the member's consolidated copy byte for byte as stored, that concurrent deletions leave no member, and that the first write through consolidated metadata stores the member's upgrade or refuses a changed grid. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * docs(group): say what consolidated metadata stores for an upgraded array Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(array): clear the stored document whenever an array stores its metadata Setting attributes of an array read from a document that had to be upgraded stored the upgraded document with the new attributes, but left the metadata marked with the document it was read from. A consolidated group handle shares that metadata, so its next write stored the old document in the consolidated metadata and the new attributes were lost from it (zarr 3.4.0 kept them). Every write of an array's own documents (creating, resizing, setting attributes) now goes through `AsyncArray._save_metadata`, which clears the mark afterwards, as the first chunk write does after storing the upgrade. Both clear it through `_stored_document_replaced`, the one place that declares it: the first chunk write clears it also when it stores nothing (the store already holds a valid document, or none), so it cannot be folded into the save. The deep copy in `mark_upgraded` stays: the metadata's attributes share objects with the document it was read from, which the caller also holds. A test pins it. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * test(metadata): pin that a failed metadata save keeps the stored document mark Setting attributes on or resizing an array read from a document that needed an upgrade clears the `_stored_document` mark only after the store accepts the new metadata. A test now injects a store failure for the array's own document and checks the mark and the stored bytes survive, for Zarr formats 2 and 3. The `_save_metadata` docstring now says only what the method does. Assisted-by: ClaudeCode:claude-opus-5-5 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(metadata): read a stored chunk size of 0 as 1, however long the axis A stored chunk size of 0 or `false` was read as one chunk spanning the axis at open time. On an axis that grew after the size was written, that made the chunk size depend on when the array was opened, and it could be very large: `chunks: [0, 100, 100]` on shape (10000, 100, 100) read as one 800 MB chunk. It is now read as the smallest valid chunk edge length (1, or the inner chunk size for a shard), which is what `chunks=-1` gives on a zero-length axis. A resize by software that kept the stored 0 no longer changes the layout a stale handle expects. Assisted-by: ClaudeCode:claude-opus-5-5 * fix(metadata): re-save only upgrades that move chunks, and read non-JSON input as is A stored `true` chunk size or an integral float edge is read as the value zarr already read it as, so chunks written under the upgraded metadata are where every reader looks. Marking such documents for re-saving turned a plain chunk write into a metadata write: in a ZipStore that adds a second `zarr.json` entry (a `UserWarning`, an error under `-W error`), and it opened a race with concurrent metadata writes that 3.4.0 did not have. Each upgrade now reports whether it moves chunks, and only those mark the metadata. The upgrades also detected changes by comparing JSON encodings, so `ArrayV3Metadata.from_dict` given metadata built in code (a codec instance, a NumPy integer in a sharding codec's `chunk_shape`) raised `TypeError`. Changes are now tracked where they are made. Assisted-by: ClaudeCode:claude-opus-5-5 * fix(array): encode new array metadata before deleting an existing node With the rectilinear flag checked when metadata is stored rather than when it is built, `zarr.create(..., overwrite=True)` with rectilinear chunks and the flag off deleted the existing array and then raised; zarr 3.4.0 raised before deleting anything. `AsyncArray._create_v2`, `_create_v3` and `init_array` now build and encode the new metadata before `_prepare_overwrite`, so metadata that cannot be stored fails with the store untouched. This also covers `create_array(..., overwrite=True)`, which deleted the existing array before validating its arguments. Assisted-by: ClaudeCode:claude-opus-5-5 * test(metadata): type the rectilinear chunks passed to zarr.create in an overwrite test Assisted-by: ClaudeCode:claude-opus-5-5 * fix(metadata): recommend recreating an array whose stored chunk size was 0 The warning for a stored chunk size of 0 on an axis of positive length now says that the array holds only its fill value, so recreating it with the wanted chunk shape loses nothing, and gives the re-save as the way to keep it instead. A re-save freezes the smallest chunk size into an array that holds no data yet. Each reading now carries its own advice, so mark_upgraded appends no shared hint. Assisted-by: ClaudeCode:claude-fable-5-1 Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> * test(metadata): widen the abbreviated-message bound for the two hints Assisted-by: ClaudeCode:claude-fable-5-1 Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> * fix(metadata): give the recipe for recreating an array read from a stored chunk size of 0 The warning names the call, zarr.from_array(array.store, name=array.path, data=array, chunks=..., overwrite=True, write_data=False), which keeps the data type, fill value, attributes and codecs, and a test pins the recipe on both formats. Assisted-by: ClaudeCode:claude-fable-5-1 Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> * test(metadata): widen the abbreviated-message bound for the recipe Assisted-by: ClaudeCode:claude-fable-5-1 Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> * refactor(metadata): call the lenient readings of stored documents repairs, not upgrades The module is zarr.core.metadata.repair: repair_array_document, mark_repaired, ARRAY_REPAIRS and the Repair type, AsyncArray._store_repaired_document, and the test module test_repair.py. A reading that turns an invalid stored document into a valid one fixes it; it does not move it to a newer format. Assisted-by: ClaudeCode:claude-fable-5-1 Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> * docs(changes): trim the #4375 fragment to what the PR adds over main Rejecting edge lists in a regular chunk grid landed with #4334, and encoding a new node's metadata before deleting the existing one landed with #4410. Assisted-by: ClaudeCode:claude-fable-5-1 Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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.
This PR fixes a nasty data deletion bug where opening / creating a group with clobber mode (mode="w" or overwrite=True) would wipe existing nodes even if something failed in creating the new node (e.g. invalid metadata).
🤖 AI text below 🤖
Opening or creating an array or group with
mode="w"oroverwrite=Trueno longer deletes the existing node when the call fails. Previously the existing node was deleted before the new one was validated, so an invalid argument (for examplezarr.open(store, mode="w", config=...), an invalidfill_value, attributes that are not JSON, or an arrayzarr.save_groupcannot store) destroyed the existing data and created nothing. Deletion now belongs to node creation:StorePath.open(mode="w")no longer erases the keys under the path, and the node written there replaces them once its metadata is valid. On a store that cannot delete keys, such asZipStore,mode="w"creates the node when the path is empty and raisesContainsArrayErrororContainsGroupErrorwhen a node already exists, instead ofNotImplementedError.Example
The agent found this while auditing
**kwargsforwarding (the same audit that produced #4408). A scan of 17 failing calls to overwriting entry points found that 12 of them lost data, includingzarr.create_array(..., overwrite=True, fill_value=<invalid>)andzarr.group(..., overwrite=True, attributes=<not JSON>).Cause
There were two deletion points, and both ran before validation:
StorePath.open(mode="w")erased the path when theStorePathwas built, before the API function dispatched. This affectedopen,open_array,open_group,create,save_arrayandsave_group.overwrite=Truedeleted the node before its metadata was built, inAsyncArray._create_v2/_create_v3,init_arrayandAsyncGroup.from_store.Fix
save_new_metadatainzarr.core.metadata.ioencodes the new metadata first, then deletes the existing node (or checks that none exists), then writes. It replaces_prepare_overwriteand the copy of that logic inAsyncGroup.from_store.StorePath.open(mode="w")no longer deletes anything; it validates the mode against the store like"a"does ("w-"still checks that the path is empty). Every API function passes its mode through unchanged, and the ones that create nodes do so withoverwrite=Truewhen the mode is"w".openandopen_arrayskip their "open the existing node" probe for mode"w".create(mode="w")now also setsoverwrite=True, which was previously implied only by the up-front erase; with aStorePathargument,mode="w"therefore now overwrites (the other modes still have no effect, as documented).save_groupchecks its arguments, and resolves every array's data type, before it replaces the group. The replacement is an explicitAsyncGroup.from_store(..., overwrite=True).NotImplementedErrorfrom the eager erase. It now takes the same path asoverwrite=Truealways did on such stores: create when the path is empty, raiseContainsArrayError/ContainsGroupErrorwhen a node exists. Themodedocstrings ofopen,open_groupandcreatesay so.A mode
"w"call that succeeds behaves as before: everything under the path is deleted and the new node is written.Tests
test_overwrite_replaces_existing_node: 10 entry points × existing array or group-with-children × zarr format 2/3, checking that a successful overwrite leaves the store exactly as the same call leaves an empty store.test_overwrite_without_deletes_creates_node/..._keeps_existing_node: the same 10 entry points overZipStore.test_overwrite_read_only_store_names_mode: the read-only error names'w'.save_groupwith a non-array argument or an array whose data type cannot be stored. Each asserts that the store contents are unchanged.test_store_path_open_w_keeps_keys:StorePath.open(mode="w")leaves keys in place.Not covered
save_groupstill replaces the group before writing its arrays, so a chunk that fails to encode while being written loses the old group. Once the arguments pass validation, that replacement is the documentedmode="w"behaviour ofsave_group.Stack bookkeeping
Rebased onto
mainafter #4455 and #4412 merged. #4455 documentedmodeoncreate; the sentence aboutStorePatharguments was updated here becausemode="w"now setsoverwrite.🤖 Generated with Claude Code