edf2csv 0.5.74 → 0.5.76

package/CHANGELOG.md

@@ -3,6 +3,46 @@
 Notable changes to edf2csv. Versions follow [semantic versioning](https://semver.org); while the
 major version is 0, a minor bump may contain breaking changes.
 
+## 0.5.76
+
+### Fixed: the reference described a line 0.4.51 had already removed, and said elsewhere that it was wrong
+
+In the `--info` section: "With `--annotations-only` the signal channels read `(not selected)`
+and the estimate is 0 rows, because that run would write no signal data."
+
+Two hundred lines further down, in the `--annotations-only` section: "Until 0.4.51 this line
+read `Would write 0 rows, roughly 0 B.`, which was true of the signal tables and false of the
+run." One page describing a behaviour and its own fix of that behaviour, disagreeing.
+
+What `--info --annotations-only` prints is neither: no row estimate at all, and a sentence
+naming the files instead — "Would write annotations.csv and channels.csv, and no signal data.
+How many events there are cannot be told from the header."
+
+The sentence now says that, and says which version it was describing, since a reader who built
+on it should be able to tell when it stopped being true. Pinned by running the command: the
+sample output the page quotes has to be what the tool prints, compared on the words rather than
+the layout, since the page wraps to its own width.
+
+## 0.5.75
+
+### Fixed: two pages gave different formulas for where an untimed record goes, and one was wrong
+
+A data record whose timekeeping annotation cannot be read has to be placed somewhere.
+`deriveRecordStarts` puts it at `origin + index * recordDuration`, where the origin comes from
+the first record that does state a time.
+
+api.md says exactly that, and says why: "not at `index * recordDuration`, which silently assumes
+the recording begins at zero". edf-plus-annotations.md then said `index * record_duration` — the
+form the other page names in order to warn against it. On `lost-timekeeping-d.edf`, whose origin
+is 0.5 s, the documented arithmetic gives 0.000 for a record the conversion writes at 0.500.
+
+Corrected, with the reason attached rather than only the formula, since the formula alone is
+what drifted. A test anchors the claim: both pages must carry the origin-aware form, and the
+sentence in edf-plus-annotations.md that states the fallback must be that one. Anchored on the
+sentence rather than on every appearance of the arithmetic — both pages also name the bare form
+in order to reject it, and flagging those would take a phrase blacklist that grows with the
+prose.
+
 ## 0.5.74
 
 ### Fixed: "1 records", "1 bytes"

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "edf2csv",
-  "version": "0.5.74",
+  "version": "0.5.76",
   "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",