edfcore 0.5.18 → 0.5.20

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 = "0.5.18";
+export declare const VERSION = "0.5.20";
 //# 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 = '0.5.18';
+export const VERSION = '0.5.20';
 //# sourceMappingURL=constants.js.map

package/docs/CHANGELOG.md

@@ -6,6 +6,50 @@ 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.
 
+## 0.5.20
+
+- **Added** the lifetime contract `api-sources.md` states in five words: of `close?()`, "edfcore
+  never calls it for you." A caller who opened a file handle, or a source wrapping a pooled
+  connection, is entitled to decide when it is released — and a library that closed it on their
+  behalf would be releasing a resource they may still be using, with the failure arriving later and
+  somewhere else. Nothing checked it. `cache.test.ts` covers the other direction, that
+  `cachedSource.close()` forwards to the source it wraps, which is the case where a caller did ask.
+- Every reading entry point that takes a source is now driven over a spy that records whether it was
+  closed: `readHeader`, `openEdf`, `inspectEdf`, `readRecordBytes`, `readRecords`, `readWindow`,
+  `readAnnotations`, `buildRecordIndex`, `validateRecording`, `streamRecords`, and one wrapped in
+  `cachedSource`. Each is checked to have actually read, so a passing run is not a source nobody
+  touched.
+- The failure paths are checked too, because a `finally` added for cleanup is exactly how this
+  contract gets broken: a file that is not an EDF at all, a read refused before it is issued, and a
+  read naming records the file does not have.
+- A behavioural sweep only covers the paths it thought to drive, so the `close` call sites are
+  enumerated out of `src/` as well. There are four, in two modules. Three are inside a `close()`
+  implementation of a source edfcore hands back; the fourth is `fileSource` releasing a handle that
+  never reached a caller, which is the exception its own docblock documents. `fileHandleSource` is
+  then checked to close its handle exactly once — when the caller calls it, and not while reading.
+
+## 0.5.19
+
+- **Added** the table on `diagnostics.md` that says which calls take `strict`. Eleven functions in
+  two columns, and nothing read it. `strict-decision.test.ts` covers what `strict` does — the info
+  exemption, the four always-fatal conditions — and `strict-reaches.test.ts` how far into a file it
+  reaches. Which entry points accept it at all was prose.
+- It is the kind of table that goes stale silently, in both directions. A function that grows a
+  `ParseOptions` parameter joins the left column without anyone editing the page, and a caller
+  reading the right column concludes it cannot be made strict. A function whose options type is
+  narrowed leaves the left column the same way, and a caller passing `strict` to it gets one that is
+  quietly ignored — an option that type-checks, runs, and does nothing.
+- The two columns are checked by different means, because they are different claims. The six on the
+  left are **driven**: each is given a file with a real non-`info` defect — a non-standard reserved
+  field for the header three, a malformed TAL for the annotation three — and each must throw
+  `EdfFormatError` carrying that code under `strict: true` and collect the same diagnostic without
+  it. That is the only way to catch a `strict` that is accepted and dropped.
+- The five on the right are checked **structurally**, out of `src/`: each function's options
+  parameter is resolved to its declared type, the `extends` and `&` chain is followed, and `strict`
+  must not be reachable through it. A behavioural check cannot prove that negative — passing an
+  option a signature does not declare is a compile error, not a runtime one. The same resolver is
+  then pointed at the left column, which is what keeps the two halves honest.
+
 ## 0.5.18
 
 - **Added** the unscalable signal `diagnostics.md` walks through, run from the throw to the read.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "edfcore",
-  "version": "0.5.18",
+  "version": "0.5.20",
   "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 = '0.5.18';
+export const VERSION = '0.5.20';