edfcore 0.4.464 → 0.4.466

package/dist/constants.d.ts

@@ -109,5 +109,5 @@ export declare const SIGNAL_FIELD_BLOCK_OFFSETS: {
     readonly reserved: 224;
 };
 /** Published package version. Kept in sync with package.json by a test. */
-export declare const VERSION = "0.4.464";
+export declare const VERSION = "0.4.466";
 //# sourceMappingURL=constants.d.ts.map

package/dist/constants.js

@@ -79,5 +79,5 @@ export const SIGNAL_FIELD_BLOCK_OFFSETS = {
     reserved: 224,
 };
 /** Published package version. Kept in sync with package.json by a test. */
-export const VERSION = '0.4.464';
+export const VERSION = '0.4.466';
 //# sourceMappingURL=constants.js.map

package/docs/CHANGELOG.md

@@ -6,6 +6,40 @@ alone does not tell you whether you were affected.
 edfcore is pre-1.0. Patch releases have carried behaviour changes where the old behaviour was a
 defect; those are called out below.
 
+## 0.4.466
+
+- **Added** tests for `buildSegmentation` over a sparse `ArrayLike`, which is what its signature
+  admits and what nothing had ever passed it. `decodeAnnotations` returns a `BigInt64Array`, where
+  an index inside the length always holds a value, so both `undefined` checks inside the walk were
+  dead: deleting either changed nothing, and the function would then throw on an input its own
+  type accepts.
+- Throwing is the wrong answer for this module. It is documented as structural — it reports the
+  shape the onsets have and judges none of it — so a hole is not a discontinuity, and a caller
+  assembling onsets from partial probes would get a stack trace instead of the segments it does
+  know about.
+- Four cases: a hole is skipped and the walk continues past it; a discontinuity between two
+  entries that ARE present still splits, so skipping a hole has not become skipping the
+  comparison; and a missing record 0 returns nothing at all, with or without an explicit origin,
+  because that onset is the axis every second here is measured from.
+- The third `undefined` check, in the gaps loop, is left alone and said so in the test: it reads
+  the `segments` array the function built itself, so no input reaches it and pinning it would mean
+  faking the array.
+
+## 0.4.465
+
+- **Added** the close half of `cachedSource`'s pass-through, which had never run. At `maxBytes: 0`
+  the function returns a different object entirely — built and returned before any cache state
+  exists — with its own `read` and its own `close`. The read half was covered twice over; the
+  close half is the one whose failure a test cannot see indirectly.
+- `cachedSource(await fileSource(path), { maxBytes: 0 })` is a reasonable thing to write. The page
+  documents the budget as floored at 0, and wrapping unconditionally while tuning the number per
+  environment is how a caller arrives there. A pass-through that dropped the delegation would hold
+  the descriptor open for the life of the process, with nothing about the reads looking wrong.
+- The optional-`close` case is pinned alongside it: `ByteSource.close` is optional and
+  `byteSource(bytes)` is the adapter that omits it, so `source.close?.()` is what makes closing a
+  zero-budget wrapper over an in-memory file resolve rather than throw a `TypeError` — the shape a
+  caller hits while swapping adapters.
+
 ## 0.4.464
 
 - **Added** tests for the `maxMaterializeBytes` boundary in all three modules that check it —

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "edfcore",
-  "version": "0.4.464",
+  "version": "0.4.466",
   "description": "Modern, typed, zero-dependency reader for EDF, EDF+, BDF and BDF+ biosignal files. Works in browsers and Node with true random access.",
   "keywords": [
     "edf",

package/src/constants.ts

@@ -93,4 +93,4 @@ export const SIGNAL_FIELD_BLOCK_OFFSETS = {
 } as const;
 
 /** Published package version. Kept in sync with package.json by a test. */
-export const VERSION = '0.4.464';
+export const VERSION = '0.4.466';