edfcore 0.5.72 → 0.5.74

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 = "0.5.72";
+export declare const VERSION = "0.5.74";
 //# 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 = '0.5.72';
+export const VERSION = '0.5.74';
 //# sourceMappingURL=constants.js.map

package/docs/CHANGELOG.md

@@ -6,6 +6,48 @@ 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.
 
+## 0.5.74
+
+- **Added** an eleventh shape to the `AWKWARD` matrix: a file with a gap AND a sub-second start
+  offset at the same time. Fifteen test files run their sweeps over that matrix, and every one of
+  them now sees it.
+- It is the fixture this project's longest-running defect class hides from. Seven separate versions
+  fixed one function deriving a time from the nominal grid while every other function used the
+  record's true onset — 0.1.4, 0.2.10, 0.2.18, 0.2.19, 0.2.28, the sample-grid family in 0.2.32,
+  and 0.2.68, where the function ADDED to fix the class shipped with the defect itself. Several of
+  them survived their own review because the fixtures had one of the two properties and never both:
+  with only a gap, `t = 0` is still the header start time and a nominal derivation looks right;
+  with only a sub-second start, the onsets are still the nominal grid and it looks right again.
+- The matrix had `EDF+D with a gap`, which starts at zero, and no shape with a sub-second start at
+  all. The new one starts at 0.25 s and skips five seconds after its third record, so `t = 0` is
+  neither the header start time nor a record boundary of the second segment.
+- Every sweep passes over it unchanged, which is the answer that was wanted and not one anybody
+  could state before. The two length assertions that pin the matrix size move from ten to eleven.
+
+## 0.5.73
+
+- **Added** the day the network hiccups. `index.onsetTicks(r)` memoises, which is why `locate()`
+  costs `O(log recordCount)` reads instead of repeating them — and memoising is the classic place
+  to store a FAILURE by accident. Cache the promise rather than the value and one dropped
+  connection is permanent: every later caller awaits the same rejection and no retry ever reaches
+  the network. Nothing tested a read that fails and is then tried again.
+- The distinction is invisible until it matters, and it is one small edit away. 0.5.44 pinned that
+  concurrent callers of `onsetTicks` all miss the memo, because it holds values and not promises —
+  and the obvious way to "fix" that is to hold the promise instead. Making exactly that change
+  fails three of the eight checks here, which is the point of writing them down.
+- `a-read-that-failed-is-not-remembered.test.ts` uses a source that rejects a stated number of
+  times and then behaves. An onset, a `locate`, and a window are each asked for while it is failing
+  and again after, and the answer after has to be the answer a source that never failed would have
+  given. The rejection arrives as the caller's own error — `isEdfError` is false for it, because a
+  dropped connection says nothing about the recording.
+- `cachedSource` gets the same question, provoked at the first read: it coalesces a small file into
+  one range, so by the time a recording is open there is nothing left for a later read to fail at.
+  A failed open leaves the cache holding nothing, and the retry through the same cache reads
+  correctly.
+- The other half is asserted so none of it passes on a package that simply never caches: a
+  SUCCESSFUL read is remembered, two calls for one record cost one read, and a record index that
+  cannot exist is refused out of the header without touching the source at all.
+
 ## 0.5.72
 
 - **Added** the case none of the four caller-supplied callbacks documents: what happens when yours

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "edfcore",
-  "version": "0.5.72",
+  "version": "0.5.74",
   "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 = '0.5.72';
+export const VERSION = '0.5.74';