Comparing releases

Comparing two versions is how you find features before they are announced.

cellar diff

With no arguments this compares the two newest versions you have stored. Name them explicitly when you want a specific pair.

cellar diff whatsapp-1030882912 whatsapp-1044822804 --filter protocol --no-hunks

whatsapp-1030882912 -> whatsapp-1044822804  (filter `protocol`)
  considered 4711 of 172040 modules (old) / 5269 of 187892 (new)
  added 913 | removed 355 | modified 1515 | unchanged 2841
  suppressed as noise: 41 | diffs skipped for size: 0

Read the summary before the details. Added and removed modules are where new features appear. Modified modules that gained dependencies or exports are where existing features grew.

Start with --no-hunks to see the shape of a release, then run it again without that flag for the handful of modules worth reading.

What counts as a change

A module is compared by the fingerprint of every definition it ships, not by its text. Reformatting alone never registers as a change, and a module that gained a second build variant does register, even if the main definition is untouched. Changes where every altered line is compiler output are counted as noiseOnly in the summary and left out of the list. Add --include-noise if you suspect the noise rules are hiding something.

Export formats

Text
JSON
NDJSON
Markdown

Human-readable, the default.

cellar diff --no-hunks

[modified] WAWebSendMsgStanza
  + deps: WAWebAddonEncryption
  81.4% similar, +4 -7 lines
  new: /Users/you/.cellar/bundles/whatsapp-1044822804/modules/WAWebSendMsgStanza.js

One document, for scripts and assistants.

cellar diff --format json | jq '.summary'

{
  "added": 913,
  "removed": 355,
  "modified": 1515,
  "unchanged": 2841,
  "noiseOnly": 41,
  "diffsSkipped": 0,
  "oldTotal": 172040,
  "newTotal": 187892
}

List just the new modules:

cellar diff --format json | jq -r '.changes[] | select(.kind=="added") | .name'

One JSON object per line. The first line is the summary with an empty change list, then one line per change, so you can stream a large comparison instead of buffering it.

cellar diff --format ndjson --no-hunks | head -2

{"changes":[],"filter":"protocol","new":"whatsapp-1044822804","old":"whatsapp-1030882912","summary":{"added":913,"modified":1515,...}}
{"kind":"added","name":"A2UIWidgetStateContext","newFile":"modules/A2UIWidgetStateContext.js","newSha256":"953e62318175bd4cd3d4039ccc41b70e049246bd6c4d729edd2d9869575065b3"}

A report you can paste into an issue or a changelog.

cellar diff --format md --limit 40 > release-notes.md

# whatsapp-1030882912 → whatsapp-1044822804

Filter: `protocol`

| metric | count |
| --- | --- |
| Added | 913 |
| Removed | 355 |
| Modified | 1515 |
| Unchanged | 2841 |
| Noise-only (suppressed) | 41 |
| Diffs skipped for size | 0 |

## added (913)

### `A2UIWidgetStateContext`

Options

Flag	Meaning
`--filter NAME`	Which modules to consider. Defaults to `default`.
`--no-hunks`	List what changed without the line-by-line diff.
`--include-noise`	Include changes that are only compiler output.
`-C N`	Context lines around each change. Default `3`.
`--limit N`	Cap the list. The summary still reports true totals.
`--platform`	Which product to pick defaults from. Defaults to `whatsapp`.

The default filter skips .pb and .graphql files, because they are generated and compare badly as text. When looking for a new protobuf field, use --filter schemas instead. See filters.

Get started

Commands

Comparing releases

What counts as a change

Export formats

Options

​What counts as a change

​Export formats

​Options

What counts as a change

Export formats

Options