edfcore 0.4.425 → 0.4.427

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 = "0.4.425";
+export declare const VERSION = "0.4.427";
 //# 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 = '0.4.425';
+export const VERSION = '0.4.427';
 //# 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.
 
+## 0.4.427
+
+- **Added** a property test for `trimToWindow` against the rule it states rather than the examples
+  it was written from. The source says which samples belong in a window in one line — "Sample j is
+  in the window when `ceil(j * D / S)` is in `[R, Rend)`" — and what the code does is a closed form
+  derived from it: two `floorDiv`s over bigint products, with the derivation written above them. A
+  closed form is exactly where an off-by-one lives, and the derivation is the part a reader takes
+  on trust.
+- So the rule is implemented the obvious way, asking every sample, and the two must select the same
+  set for arbitrary geometries and windows. The naive version is too slow for a library and is
+  obviously right, which is the only pairing worth testing a closed form against.
+- The rounding it turns on is not incidental. 256 samples in a one-second record puts sample 1 at
+  39,062.5 ticks, published as 39,063, and selecting on the exact start rather than the published
+  one excluded that sample from a window beginning at its own published start — half of all indices
+  at that rate, and at 128 samples per 0.29 s a one-sample window came back empty (0.3.56).
+  Geometries whose boundaries miss whole ticks are generated deliberately, because the ones that
+  land on them cannot tell the two rules apart.
+- Three obligations from the docblock come with it, none previously checked in general: adjacent
+  windows partition a chunk exactly, a window covering the chunk is the identity, and the result is
+  a view rather than a copy.
+
+## 0.4.426
+
+- **Added** a property test that the cache is invisible at every size it can be configured to.
+  `api-sources.md` describes `cachedSource` as "removed by deleting one wrapper from the expression
+  that built the source", which is the property a caller relies on when they add it: the reads get
+  cheaper and nothing else changes.
+- The existing demonstration is one script of six reads at the default block size, where the 1 MiB
+  block swallows the fixture whole — so it demonstrates a cache that never evicts, never stitches
+  and never splits. The interesting sizes are the other ones: a block smaller than a read makes
+  every answer a stitch, a budget smaller than a few blocks makes the cache evict mid-sequence, and
+  a read wider than the budget bypasses the cache entirely. All three are reachable from an
+  ordinary configuration — blocks sized to a record, a budget sized to a phone.
+- The failure is not a crash. Stitching arithmetic off by a block start returns the right number of
+  bytes from the right file, taken from the wrong offset: a header that parses, samples that plot,
+  and a recording quietly shifted.
+- The copy rule is scoped rather than assumed. It is asserted for reads the cache actually serves;
+  a read wider than the whole budget returns the wrapped source's own array by design, because
+  nothing is retained on that path, and `cached.ts` says so. Each property builds its fixture
+  fresh, since a test that writes into a result would otherwise edit the bytes it compares against
+  and pass on the damage.
+
 ## 0.4.425
 
 - **Added** tests for the one sentence `api-reading.md` gives about the shape of `signalIndices`:

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "edfcore",
-  "version": "0.4.425",
+  "version": "0.4.427",
   "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 = '0.4.425';
+export const VERSION = '0.4.427';