edfcore 0.6.35 → 0.6.37

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 = "0.6.35";
+export declare const VERSION = "0.6.37";
 //# 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 = '0.6.35';
+export const VERSION = '0.6.37';
 //# sourceMappingURL=constants.js.map

package/docs/CHANGELOG.md

@@ -6,6 +6,53 @@ 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.
 
+## 0.6.37
+
+- **Fixed** an incomplete sentence on `api-reading.md`. It said the index traversal "is chunked so
+  memory stays bounded whatever the file size: each chunk holds
+  `min(4 MiB, maxMaterializeBytes) / recordByteLength` records, and never fewer than one" — which
+  is true and reads as though a small budget only makes the scan slower. Below one record's worth
+  it does not: one record cannot be read within the budget and there is no smaller count to ask
+  for, so the scan is refused. The page says that now, and says the refusal names the record's size.
+- **Added** `the-page-says-always.test.ts`, which executes the five claims on that page that say
+  "always", "never" or "every" against all seventeen shapes. Each was written when the matrix held
+  eight, and a universal claim is only as checked as the widest file anyone pointed it at — 0.6.36
+  is what happens when one meets a shape it had not: four sweeps turned out to be asserting
+  something narrower than they said. The claims are read off the page, so a rewording fails here.
+- Two are worth naming. `readAnnotations` answers on a file with NO annotations channel, returning
+  one onset per record from the nominal grid — "always" includes the file with nowhere to store an
+  annotation. And on that same file a budget of one byte does not refuse, because the page's next
+  sentence says such a file "is not scanned at all, because its record onsets are arithmetic": the
+  two claims agree, and the test checks that they do rather than exempting one.
+- No library change.
+
+## 0.6.36
+
+- **Added** a seventeenth shape to the `AWKWARD` matrix: records that overlap in time, where a
+  record starts before the one before it ended and two records claim the same instant. edfcore has
+  always known about them — the gap list carries one with a negative duration (0.2.69),
+  `mergeChunks` refuses the join, `edfcore gaps` prints it in its own column, and the probe raises
+  `RECORD_ONSET_SPACING_VIOLATION`, a warning rather than an error because an overlapping file is
+  still readable.
+- **Fixed** four sweeps that were asserting things true only of files whose records do not overlap,
+  which nothing had ever supplied:
+  - `header-helpers-agree` said the sum of the record durations never exceeds the span. It exceeds
+    it by exactly the overlap, because an overlap counts an instant twice.
+  - `timeline-helpers-agree` said an instant inside a segment is answered with THAT segment, and
+    that a gap's midpoint is in no segment. Two segments contain an overlapped instant, and an
+    overlap's interval runs backwards, so its midpoint is inside a segment rather than in a hole.
+  - `reading-helpers-agree` said a window at a sample's start begins at that sample's record, and
+    that a window's chunks join. Neither holds here — and the second was the half of its own
+    comment the code had never run: "either join or are refused with a reason" had only ever seen
+    joining.
+- None of the four was a defect in `src/`. The library refuses the unindexed read with a message in
+  the file's own numbers ("records covering 6 s are packed into a 3.5 s span"), returns both
+  covering records in separate chunks once it has an index, and refuses to concatenate them. Each
+  sweep now states the weaker thing true of every file and the stronger thing true of this one,
+  rather than exempting the shape.
+- That is what a matrix is for: the claims were not wrong because anyone was careless, they were
+  claims with no counterexample.
+
 ## 0.6.35
 
 - **Added** the census that closes the class four releases in a row have been fixing. A sample rate

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "edfcore",
-  "version": "0.6.35",
+  "version": "0.6.37",
   "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 = '0.6.35';
+export const VERSION = '0.6.37';