edfcore 0.5.35 → 0.5.37

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 = "0.5.35";
+export declare const VERSION = "0.5.37";
 //# 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 = '0.5.35';
+export const VERSION = '0.5.37';
 //# sourceMappingURL=constants.js.map

package/docs/CHANGELOG.md

@@ -6,6 +6,55 @@ 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.
 
+## 0.5.37
+
+- **Added** the safety property for the CLI. `fuzz.test.ts` states it for the library — for any byte
+  sequence, edfcore either parses it or throws an `EdfError`, never a bare `TypeError`, never a
+  hang, never believable garbage. The CLI is a second surface over the same parser, with its own
+  formatting, six commands and exit codes, and nothing had ever pointed it at bytes nobody chose.
+  Every CLI test in the suite feeds it a file written to make a point.
+- That matters because of how `cli.ts` ends. `main()` catches everything and reports
+  `edfcore: ${error.message}` with exit 1, so a `TypeError` escaping `runCli` does not crash — it
+  arrives as a line with no code, no byte offset and no `Next:` clause, wearing the prefix a real
+  diagnostic wears. The failure is indistinguishable from a working refusal unless something checks
+  the class.
+- Two properties now run over ~1,300 cases — random bytes, single-bit flips of a good file,
+  truncation at every length, and each header and per-signal field replaced with something that
+  breaks it. Every rejection is an `EdfError` whose message opens with a bracketed code and ends
+  with a `Next:` clause. And nothing reaches stdout before a rejection, so
+  `edfcore json big.edf > out.json` either writes a whole document or writes nothing — a redirect
+  cannot leave a half-written file beside a non-zero exit.
+- The exit codes are checked for reachability rather than assumed: 0 for a clean file, 1 for one
+  that reads and fails validation, 2 for bad usage. `parseArgs` over 500 arbitrary argv arrays
+  throws nothing but `CliUsageError`, which is what keeps 2 distinguishable from 1.
+
+## 0.5.36
+
+- **Added** the `ByteSource` `data-sources.md` tells you to write, written and read through. "Writing
+  your own ByteSource" is the section that makes the interface a promise rather than an
+  implementation detail — "the whole job is: know the length, return exactly the bytes asked for, and
+  reject otherwise" — and it prints a complete forty-line adapter over a range-returning object
+  store, with four rules under it "in order of how badly they bite". None of it was run.
+  `source-contract.test.ts` covers the contract from edfcore's side, with sources written for the
+  test; the adapter a reader would copy had nothing.
+- It is now transcribed and used: a recording is opened over it, a window is read, and the samples
+  are compared against the same file through `byteSource`. The store streams in irregular pieces —
+  7, 1, 4096, 3 — which is the whole reason the adapter has a loop, and a test that returned each
+  range in one go would never reach it. That is asserted too: no single piece the store hands back
+  is a whole read.
+- Each of the four rules is then broken on purpose, in an adapter of its own. A short return is
+  caught by the adapter, which can name the key and the range, and by edfcore anyway for a source
+  that does not check. A padded one is the dangerous case: on a plain EDF every sample of every
+  record comes back as a legal zero with no diagnostic at all, and only the annotation region of an
+  EDF+ notices, with `TIMEKEEPING_TAL_MISSING`. A reused buffer is shown corrupting an array a
+  caller had already been handed, which the page's own per-read allocation does not.
+- The fourth is the page's Warning, and it is the one worth having under test: "edfcore itself does
+  not poll the signal between reads. A custom `ByteSource` that ignores `options.signal` makes
+  cancellation a complete no-op, including for a long `validateRecording` sweep that issues hundreds
+  of reads." A whole-file sweep and a chunked index traversal both run to completion over an
+  already-aborted signal, serving twenty reads after the abort. A stated limitation nobody checks is
+  indistinguishable from a bug.
+
 ## 0.5.35
 
 - **Added** the resolution of every name a `Next:` clause points at. `next-clause.test.ts` proves

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "edfcore",
-  "version": "0.5.35",
+  "version": "0.5.37",
   "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 = '0.5.35';
+export const VERSION = '0.5.37';