edf2csv 0.4.76 → 0.4.78

package/CHANGELOG.md

@@ -3,6 +3,48 @@
 Notable changes to edf2csv. Versions follow [semantic versioning](https://semver.org); while the
 major version is 0, a minor bump may contain breaking changes.
 
+## 0.4.78
+
+### Fixed: `sleep-study.edf` was four different recordings across the site
+
+The same filename appears on the landing page, in getting-started, in cli-reference and in
+edf-format, and each showed `--info` output for a different file: 28800 records of EDF+ at
+100/10/1 Hz, three records of plain EDF at 256/128/1 Hz, 29550 records at 25.1 MB, and three
+records carrying a real patient identifier. A reader working through the pages in order was
+told the name meant something different each time.
+
+Worse than untidy in one place: getting-started showed a three-second recording, then said
+"on a long recording, `--info` tells you four things before you spend any disk", then told
+the reader to run `--start 30m --duration 5m` on it — a window that file does not have.
+
+The two channel-table blocks are regenerated from the recording 0.4.77 made reproducible, so
+they are that file's real output. The identification-fields example in edf-format is a
+different recording and now has a different name, `telemetry-psg.edf`, with a line saying
+why: it carries the EDF+ specification's own example patient header, and `sleep-study.edf`
+is anonymised, so it cannot demonstrate what that section is about.
+
+The test from 0.4.77 now checks every page: any block that opens with `File       sleep-study.edf`
+must match what the tool prints for that recording.
+
+## 0.4.77
+
+### Fixed: the landing page showed an annotations.csv the recording could not produce
+
+The page says every terminal block and CSV sample on it is real output from one synthetic
+sleep recording, and the recipe for that recording lived in a comment. Which made the claim
+unfalsifiable in practice: checking it meant rebuilding a 19 MB file by hand from prose. The
+page drifted twice anyway — a row figure in 0.4.67, a `UTC` suffix in 0.4.68 — and this
+found a third. Its `annotations.csv` sample shows `Sleep stage W` at onset 0. The recording
+had an annotation channel with no events in it at all, so the real file is a header row and
+nothing else.
+
+The recipe is a module now, `test/fixtures/sleep-study.mjs`, and the recording carries the
+sleep staging a sleep study carries. A test rebuilds it, runs `--info`, and compares the
+output against the page character for character; then converts two seconds of it and checks
+each CSV sample really is how that file starts. Samples that are abridged say so with an
+ellipsis and are left alone — which the `metadata.json` sample did not, while showing
+`"version": "0.1.0"`. It is now visibly an extract.
+
 ## 0.4.76
 
 ### Added: the API reference's examples are run, not just displayed

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "edf2csv",
-  "version": "0.4.76",
+  "version": "0.4.78",
   "description": "Convert EDF, EDF+ and BDF biosignal recordings (European Data Format) to CSV from the command line. Local, streaming, and never resamples or alters units.",
   "keywords": [
     "edf",