edfcore 0.5.31 → 0.5.33

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 = "0.5.31";
+export declare const VERSION = "0.5.33";
 //# 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 = '0.5.31';
+export const VERSION = '0.5.33';
 //# sourceMappingURL=constants.js.map

package/docs/CHANGELOG.md

@@ -6,6 +6,57 @@ 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.
 
+## 0.5.33
+
+- **Added** an overlap driven through every surface that could call it a gap. This is the defect the
+  changelog has fixed four times. An overlap travels in `index.gaps` with a NEGATIVE duration —
+  0.2.69 decided that and `api-reading.md` documents it — so every consumer of that array has to
+  branch on the sign, and each one that did not printed "a gap of -0.2 s": a gap of negative
+  duration, with an explanation that inverts what an overlap does. Across a gap two samples are
+  seconds apart; across an overlap they cover the same time, so concatenating duplicates it rather
+  than skipping it.
+- 0.3.3 partitioned `edfcore gaps`. 0.3.33 applied the rule to "the two places that still said it
+  was". 0.3.41 found a third in `src/chunks.ts`, which mentioned an overlap nowhere. 0.3.59 found a
+  fourth forty lines below the third, on the branch a probed index actually reaches. Every fix was
+  local, and no test treated it as one rule over one file.
+- One overlapping recording now goes through all of them — both `mergeChunks` branches, the
+  validation report on an EDF+D and on an EDF+C that claims to be continuous, the CLI's summary line
+  and its tab-separated row, and the record probes at open — and each is asserted to use the word
+  `overlap` and to present no negative magnitude as a duration. The second is what would have failed
+  all four times: `gap of -0.2 s` is what the defect looked like on screen, at every one of them.
+- A blanket ban on the word "gap" would be wrong, and the test says so: the diagnostics correctly use
+  it in the rule they cite — "a discontinuous file may leave gaps between records but never overlaps
+  them" — which is the sentence that makes them understandable. The check is about how the observed
+  boundary is described, not about a vocabulary.
+- The three modules that partition a gap's own duration by its sign are read out of `src/`, so a
+  fourth consumer fails here until it is driven. That is the part that stops this being a fifth
+  local fix.
+
+## 0.5.32
+
+- **Added** a lattice check over the ten microvolt values `quick-start.md` prints. That page prints
+  two `Float64Array` blocks — five values from a window at `t = 0` in the browser example, five more
+  from `t = 60` in the Node one. They are the first numbers anyone sees from this library, and
+  `quick-start-page.test.ts` runs everything around them: the header listing, the sample count, the
+  byte count, the annotation lines, the three refusals. The values themselves it did not touch.
+- They cannot be reproduced from the page, which names its file's geometry and not its waveform.
+  What they can be checked against is the scale that geometry implies, and that turns out to be the
+  check worth having: a physical value is `bitValue * (offset + digital)` for an integer `digital`,
+  so on a given declaration the reachable values are a lattice, and a float that is not on it was
+  produced by a different expression or a different range.
+- That is not hypothetical. 0.5.6 found exactly that on `api-types.md` — two printed values belonged
+  to a file with a different digital range from the one the rest of the page's example used, and the
+  existing test checked five of seven lines and stopped before them. A lattice check catches it
+  without needing the waveform, and would have caught it there. The last test makes the point
+  concretely by converting the same digital sample on a 12-bit declaration and showing it lands
+  somewhere else.
+- Each of the ten is inverted through the declared scale, asserted to be a whole number inside the
+  declared range, and put back through `toPhysical` on a real signal, compared with `Object.is` —
+  `toBeCloseTo` would pass on a value from the wrong scale. The page's own argument for the array
+  type goes with them: every one of the ten changes under `Math.fround`, so "a sample scaled into
+  float32 loses about a quarter of a quantisation step" is true of the very values printed beneath
+  it. And the first value is asserted to be what digital zero converts to, which is not zero.
+
 ## 0.5.31
 
 - **Added** the record-range contract at the five calls that take one. `RecordRange` is the other

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "edfcore",
-  "version": "0.5.31",
+  "version": "0.5.33",
   "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 = '0.5.31';
+export const VERSION = '0.5.33';