edfcore 0.5.25 → 0.5.27

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 = "0.5.25";
+export declare const VERSION = "0.5.27";
 //# 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 = '0.5.25';
+export const VERSION = '0.5.27';
 //# sourceMappingURL=constants.js.map

package/docs/CHANGELOG.md

@@ -6,6 +6,52 @@ 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.
 
+## 0.5.27
+
+- **Added** a non-finite bound driven through every entry point that takes a time in seconds.
+  `secondsToTicks` is the one place a caller's seconds become the integers everything else works in,
+  and it refuses a non-finite one; `ticks.test.ts` covers that, once, on the resolver.
+- The resolver is not the guarantee. The guarantee is that a `NaN` cannot get past it, and `NaN`
+  reaches a time bound the way it reaches a byte budget: `Number(searchParams.get('t'))`, a slider
+  whose value has not been set, an absent key in a saved view. What makes it worth a sweep is what
+  happens if one entry point misses. `NaN` compares false against everything, so a window bound that
+  skipped the resolver would not throw — it would return `[]`, which is the legitimate answer for a
+  window past the end of a recording and indistinguishable from one. That contrast is asserted
+  rather than described.
+- The entry points are enumerated out of `src/` the way `nan-budget-reaches.test.ts` enumerates the
+  budget's: sixteen exported functions declare one of `seconds`, `startSeconds`, `durationSeconds`
+  or `secondsPerBucket`, directly or through a selection type whose `extends` chain is followed.
+  Two of the sixteen take a `startSeconds` edfcore itself produced — `envelopeOfSamples` and
+  `toPhysicalEnvelope`, on an `EdfChunkSignal` and an `EdfEnvelopeSignal` — and that is stated
+  rather than filtered out silently.
+- Each of the remaining thirteen, plus `index.locate`, is driven with `NaN`, `Infinity` and
+  `-Infinity`, and each must throw a plain `RangeError` — `isEdfError` false, since a bad bound is a
+  caller's mistake and never a file's — whose message names the value it was given and ends with a
+  `Next:` clause.
+
+## 0.5.26
+
+- **Added** the two sentences that promise nothing changes when you run it again. `validation.md`:
+  "It does not modify the recording, the header or the source. `validateRecording` reads and
+  returns; call it twice and you get the same report." `diagnostics.md`: the formatter output "is
+  deterministic ... so it's safe to snapshot in a test." Both are properties of every read in the
+  package rather than of one call, and neither had a test — nothing in the suite called anything
+  twice and compared, and nothing checked that a read leaves the header it was given alone.
+- They are easy properties to lose and hard to notice losing. A parse that memoises a derived value
+  onto the header it returns, a formatter that sorts its input in place, a scan that fills in
+  `segments` on the index it was handed: each is a reasonable-looking optimisation, each makes the
+  second call disagree with the first, and none fails anything until a caller compares two runs. The
+  failure then looks like a file that changed.
+- So both are swept over the whole shape matrix rather than asserted on one fixture. For each of the
+  eight `AWKWARD` files — a zero record duration, no data signal at all, duplicate labels, a signal
+  with no usable scale — every read and every formatter runs twice and the results are compared; the
+  source's bytes are compared before and after; and the header, timeline and index are compared
+  through a walk that handles the `bigint`s and typed arrays `JSON.stringify` refuses.
+- The stronger form goes with it: a second recording opened over a copy of the same bytes must agree
+  with the first, so the output is a function of the file rather than merely stable per object. And
+  the probed index stays probed — `buildRecordIndex` returns a new one rather than filling in the
+  one it was handed.
+
 ## 0.5.25
 
 - **Added** the three option tables on `api-helpers.md` and the defaults they promise. Seven fields

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "edfcore",
-  "version": "0.5.25",
+  "version": "0.5.27",
   "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 = '0.5.25';
+export const VERSION = '0.5.27';