Give every blob value the full { data, uri } shape before inference so rows
mixing bytes, URIs, and partial structs infer one struct type. Also wrap
blob validation errors with the field and row, regenerate the TypeScript
docs, and cover URL handling end to end.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012U4KdHofT5edv3LcXHXbLv
Blob v2 columns now accept ArrayBuffer (wrapped without copying), URL
(stored as its href), and Blob/File (read with arrayBuffer() in the async
createTable/add/mergeInsert paths), both as the column value and inside
{ data } / { uri }. BlobInput is widened and exported, along with BlobData
and BlobUri, so rows can be annotated and checked by tsc.
Closes#4270
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012U4KdHofT5edv3LcXHXbLv
## Cause
Schema inference treated unrecognized objects as nested structs. Objects
without enumerable fields produced no field paths, so schema matching
dropped their columns before Arrow or blob conversion. Blob instances
with own fields were mistaken for struct fields.
## Fix
Walk only nonempty plain records as structs. Reject unsupported object
values in scalar fields, nested structs, and list elements with the
field path and row number. Preserve the object types already passed
through to Arrow. Support for converting these new input types is
tracked in #4270.
Nested class instances with enumerable fields previously became structs;
this change rejects them and is labeled `breaking-change`.
## Validation
- `pnpm build`
- `pnpm lint`
- `pnpm test __test__/arrow.test.ts __test__/blob.test.ts
__test__/table.test.ts --runInBand --silent` (626 passed)
- `pnpm run docs`
Fixes#4269
<!-- lance-gatekeeper-fix:v1 agent=426011291c209da027689ae88a0b4da7
generation=1 -->
---------
Co-authored-by: Gatefixer <313497061+lancedb-gatefixer[bot]@users.noreply.github.com>
this PR blob v2 field helpers and reads to the Node SDK.
`blob()` marks a field as blob v2 and lets you set the storage
thresholds. Inputs can be bytes, a URI, or a data/uri struct.
Queries return descriptors. `fetchBlobs()` reads the bytes by row ID,
and `fetchBlobFiles()` gives you lazy handles for full or range reads.
`blobColumns()` lists the blob fields, including nested ones.
Fetch uses the table’s current checkout. It preserves order, duplicates,
and nulls. Holding row IDs across compaction still requires stable row
IDs.
```javascript
const db = await connect("./data");
const video = await readFile("clip.mp4");
const table = await db.createTable(
"videos",
[{ id: 1n, video }],
{
schema: new Schema([
new Field("id", new Int64()),
blob("video"),
]),
},
);
const rows = await table.query().select(["id"]).withRowId().toArray();
const rowIds = rows.map((row) => row._rowid as bigint);
const bytes = await table.fetchBlobs("video", rowIds);
const [handle] = await table.fetchBlobFiles("video", rowIds);
const header = await handle!.readRange(0n, 65536n);
```
### Testing
- cover input validation, thresholds, nested fields, fetch ordering,
nulls, and range reads.
Claude
lancedb-gatefixer[bot]
Drew