edfcore 0.5.57 → 0.5.59
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 +42 -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 = "0.5.
|
|
112
|
+
export declare const VERSION = "0.5.59";
|
|
113
113
|
//# sourceMappingURL=constants.d.ts.map
|
package/dist/constants.js
CHANGED
package/docs/CHANGELOG.md
CHANGED
|
@@ -6,6 +6,48 @@ 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
|
+
## 0.5.59
|
|
10
|
+
|
|
11
|
+
- **Added** the round trip a streaming pipeline actually performs: stream a window in pieces, merge
|
|
12
|
+
the pieces, and get the window. `merge-split.test.ts` proves it for `readWindow`. `streamRecords`
|
|
13
|
+
is the other way to get a window in pieces and is different code — a `readRecords` per piece, a
|
|
14
|
+
seam check between them, and a `chunkRecords` knob that decides how many pieces there are. Its
|
|
15
|
+
docblock says the pieces must be "the same object in every respect" as a read chunk, and
|
|
16
|
+
`stream-equals-read.test.ts` checks that one piece at a time. Whether they reassemble was not
|
|
17
|
+
checked.
|
|
18
|
+
- If they do not join, the memory bound the call exists for is bought with an answer the caller
|
|
19
|
+
cannot reconstruct — which is the whole point of streaming a twelve-hour recording and then
|
|
20
|
+
running a filter over it.
|
|
21
|
+
- The stream is merged at four chunk sizes, from one piece per record to a single piece, and
|
|
22
|
+
compared against `readWindow` on the same window: the samples, the record range, the byte offset
|
|
23
|
+
and length, the exact start and duration in ticks, `precededByGap` and the diagnostics.
|
|
24
|
+
- On a discontinuous file the answer changes, correctly. The stream crosses the gap in pieces, so
|
|
25
|
+
merging all of them is refused with the message `chunks.ts` produces — and the advice that refusal
|
|
26
|
+
ends with is followed: grouped by run, each run reassembles into exactly the chunk `readWindow`
|
|
27
|
+
returns for it.
|
|
28
|
+
|
|
29
|
+
## 0.5.58
|
|
30
|
+
|
|
31
|
+
- **Added** the composition rule for `trimToWindow`: trimming to A and then to B is trimming once to
|
|
32
|
+
their overlap. The call returns a `subarray` view rather than a copy — "so trimming allocates
|
|
33
|
+
nothing and the two share storage" — which makes its result a chunk signal like any other, and a
|
|
34
|
+
caller who has one narrows it again as a matter of course: a viewer that has read a minute and is
|
|
35
|
+
now drawing ten seconds of it does exactly that.
|
|
36
|
+
- `trim-window.test.ts` covers the rule for one trim, against the closed form the source derives.
|
|
37
|
+
Composing two was not covered, and it is where a narrowing that treated its input's
|
|
38
|
+
`firstSampleIndex` as zero would show — the second trim measured from the wrong origin, coming
|
|
39
|
+
back with the right number of samples from the wrong place. The fixture's samples are a ramp over
|
|
40
|
+
the whole recording, so a sample's value names its position and a wrong origin is a different
|
|
41
|
+
number rather than a different length.
|
|
42
|
+
- The property runs over 300 arbitrary window pairs with a constant seed, including pairs that do
|
|
43
|
+
not nest and pairs that miss each other entirely.
|
|
44
|
+
- **Found while writing it,** and now stated rather than asserted away: an empty trim has no
|
|
45
|
+
position to agree about. The docblock promises a window that misses the chunk "yields a
|
|
46
|
+
zero-length result rather than an error" and says nothing about where that result sits, so two
|
|
47
|
+
routes to an empty answer report different `firstSampleIndex` and `startSeconds` — each being
|
|
48
|
+
where its own arithmetic stopped. `sampleCount` is the field to branch on, and the property
|
|
49
|
+
compares positions only where there are samples in them.
|
|
50
|
+
|
|
9
51
|
## 0.5.57
|
|
10
52
|
|
|
11
53
|
- **Added** the property that makes an envelope exhaustive: the buckets are a partition. `readEnvelope`
|
package/package.json
CHANGED
package/src/constants.ts
CHANGED