edfcore 0.5.43 → 0.5.45

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 = "0.5.43";
+export declare const VERSION = "0.5.45";
 //# 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 = '0.5.43';
+export const VERSION = '0.5.45';
 //# sourceMappingURL=constants.js.map

package/docs/CHANGELOG.md

@@ -6,6 +6,54 @@ 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.
 
+## 0.5.45
+
+- **Added** the relationship between `inspectEdf` and the call it is triage for. `diagnostics.md`
+  calls it "the first call" for an unfamiliar file: at most 128 KiB, never throws about content,
+  returns `ok`, `variant`, `header`, `bytesRead` and a diagnostics list. Every one of those is
+  tested — `inspect-safety.test.ts` over bytes nobody chose, `inspect-validate.test.ts` on the
+  ceiling — and the relationship to `openEdf` was not.
+- That relationship is the whole point. Running `inspectEdf` over a directory decides which files
+  are worth opening, and the decision is worthless if the header it shows you is not the header you
+  will get. Over the eight `AWKWARD` shapes, wherever it returns a header it is now compared field
+  for field against `openEdf`'s — diagnostics and variant included.
+- `ok` is checked as the rule the page states rather than as a value: "true only when the header
+  parsed **and** carried no error-severity diagnostic", verified against `summarizeDiagnostics` over
+  the same header, with both branches reached in the matrix. A signal with no usable scale gives
+  `ok: false` on a header that is perfectly readable and a file that opens — the distinction the
+  callout on the same page draws.
+- `variant` is checked as the "separate best effort" the page says it is. The version block and the
+  reserved field are the first 8 and 44 bytes and "stay readable long after everything else has
+  stopped making sense": a file whose signal count is garbage has no header at all and is still
+  reported as `BDF+C`, or `EDF+D`, rather than as nothing. That is the sentence the section ends
+  with, and nothing had run it.
+- The cost is bounded from both sides: never above the 128 KiB ceiling, and never above what opening
+  the same file costs, in bytes and in reads — because it skips the record probes.
+
+## 0.5.44
+
+- **Added** reading one recording from several places at once. Everything in the suite reads
+  sequentially, and nothing in the API documents an ordering requirement — a recording is a plain
+  struct, `readWindow` takes it and returns a value, and the obvious thing to write in a viewer is
+  `Promise.all` over the channels or the visible range. That was untested territory: the reads share
+  a source, an index whose `onsetTicks` memoises, and a timeline every rebasing path consults.
+- Correctness holds. Five overlapping windows resolved together give what the same five give one at
+  a time, and so does the whole API at once — a window, a record range, the annotations, an index
+  build and a full validation sweep, launched together and compared against the same five in order.
+  It also holds over a source that resolves reads in **reverse order of arrival**, which is the case
+  worth having: a network returns what it returns when it returns it, and a reader that assumed its
+  own issue order would come apart exactly there.
+- Then the one place concurrency costs something, which is worth naming because it is invisible.
+  `record-index.ts` says `onsetTicks(r)` "reads that ONE record and memoises the answer", and
+  `locate-cost.test.ts` checks that a second call is free. It memoises on **resolution** rather than
+  on request, so ten calls for the same record launched together issue ten reads. The eleventh,
+  after they settle, issues none.
+- That is a cost rather than a defect, and the package already documents the remedy: `cachedSource`
+  exists so that "concurrent reads wanting the same block issue ONE underlying read". Wrapped in it,
+  the same ten calls issue none at all — the block was already resident from the open, which is the
+  observation `large-files.md` makes about the reads that come with opening a file not being wasted.
+  The answer is identical either way.
+
 ## 0.5.43
 
 - **Added** the two halves of what a sample width decides, written side by side. EDF stores each

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "edfcore",
-  "version": "0.5.43",
+  "version": "0.5.45",
   "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 = '0.5.43';
+export const VERSION = '0.5.45';