edfcore 0.4.415 → 0.4.417
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/constants.d.ts +1 -1
- package/dist/constants.js +1 -1
- package/docs/CHANGELOG.md +32 -0
- package/package.json +1 -1
- package/src/constants.ts +1 -1
package/dist/constants.d.ts
CHANGED
|
@@ -109,5 +109,5 @@ export declare const SIGNAL_FIELD_BLOCK_OFFSETS: {
|
|
|
109
109
|
readonly reserved: 224;
|
|
110
110
|
};
|
|
111
111
|
/** Published package version. Kept in sync with package.json by a test. */
|
|
112
|
-
export declare const VERSION = "0.4.
|
|
112
|
+
export declare const VERSION = "0.4.417";
|
|
113
113
|
//# sourceMappingURL=constants.d.ts.map
|
package/dist/constants.js
CHANGED
package/docs/CHANGELOG.md
CHANGED
|
@@ -6,6 +6,38 @@ alone does not tell you whether you were affected.
|
|
|
6
6
|
edfcore is pre-1.0. Patch releases have carried behaviour changes where the old behaviour was a
|
|
7
7
|
defect; those are called out below.
|
|
8
8
|
|
|
9
|
+
## 0.4.417
|
|
10
|
+
|
|
11
|
+
- **Fixed** the cost of `spec-references.test.ts`, which was the most expensive file in the suite by
|
|
12
|
+
a wide margin. Seven checks in it ask questions of the same set of diagnostics, and each rebuilt
|
|
13
|
+
it from scratch: nine targeted parses plus 2,700 bit-flipped ones, seven times over. The sweep is
|
|
14
|
+
now run once for the file. It drops from 13.0 s to 4.4 s, and the work inside the tests from
|
|
15
|
+
10.3 s to 1.9 s.
|
|
16
|
+
- The reason to care is not the seconds. Every one of those seven checks sat within reach of
|
|
17
|
+
vitest's default 5 s timeout, which makes a test that passes or fails on how busy the machine is —
|
|
18
|
+
and two of them did fail, on a laptop running a video call and an emulator, for reasons that have
|
|
19
|
+
nothing to do with the code. A shared CI runner is the same machine on a bad day, and
|
|
20
|
+
`scripts/release.mjs` gates the tag on CI going green.
|
|
21
|
+
- Memoised rather than hoisted to a module-level `await`, so the sweep starts when the first check
|
|
22
|
+
asks for it and a rejection surfaces inside a test rather than as an unhandled one.
|
|
23
|
+
|
|
24
|
+
## 0.4.416
|
|
25
|
+
|
|
26
|
+
- **Added** tests for a file with no annotations signal, whose record onsets are arithmetic. Plain
|
|
27
|
+
EDF and plain BDF carry no timekeeping TALs, so record `r` starts at `r * recordDuration` by
|
|
28
|
+
definition and reading the data would answer a question the bytes do not contain.
|
|
29
|
+
- The consequence is a cost, and cost is why the index is shaped the way it is.
|
|
30
|
+
`locate-cost.test.ts` pins the EDF+ numbers the page prints for a file whose onsets have to be
|
|
31
|
+
read. For the majority of files in the world the number is zero, and nothing said so — a refactor
|
|
32
|
+
that probed unconditionally would be invisible in every result, because the probe would derive
|
|
33
|
+
the same arithmetic value it found written nowhere. Only the request count changes, and on a
|
|
34
|
+
remote recording that is the difference between opening a file and paying one range request per
|
|
35
|
+
step of a binary search.
|
|
36
|
+
- `buildRecordIndex` is the sharper case. On an EDF+ file it is a full traversal, which is why the
|
|
37
|
+
page tells you to gate it on the two-probe verdict; on a plain EDF it must be free, and it still
|
|
38
|
+
has to call `onProgress` once with the traversal complete, so the fastest possible file is not
|
|
39
|
+
the one whose progress bar never moves.
|
|
40
|
+
|
|
9
41
|
## 0.4.415
|
|
10
42
|
|
|
11
43
|
- **Raised** the test-count figure from 2,000 to 2,500 in the README's status line, the note at the
|
package/package.json
CHANGED
package/src/constants.ts
CHANGED