Idea
A way to inspect the baseline debt: which baselined files are worth pulling back into the gate, ranked by what optimizing them would actually buy. Something like glimpse baseline --audit or glimpse analyze --include-baselined.
Why
The baseline mixes two kinds of entries: files glimpse actually worked on (via: convert|optimize|resize|thumbnail) and files that were bulk-accepted on day one (via: analyze). The via field (#24) tells you why an entry exists, but not what it is worth: a bulk-accepted file at 3% estimated savings is fine where it is, while one at 85% is the debt worth burning down.
What it should do
- Re-analyze baselined files (they are normally skipped) and report estimated savings per file, sorted biggest saver first, like the
analyze <dir> table.
- Show each entry's
via, so accepted debt and processed files are distinguishable at a glance.
--json output for scripting.
- Read-only: it must not touch the baseline; acting on a finding is still
optimize/convert (which re-record automatically) or analyze --update-baseline.
Open questions
- New subcommand (
glimpse baseline ..., leaving room for future baseline tooling) vs a flag on analyze?
- Should a
--min-savings threshold filter the noise?
Came out of the review discussion on #24, alongside #25 (glimpse init).
Idea
A way to inspect the baseline debt: which baselined files are worth pulling back into the gate, ranked by what optimizing them would actually buy. Something like
glimpse baseline --auditorglimpse analyze --include-baselined.Why
The baseline mixes two kinds of entries: files glimpse actually worked on (
via: convert|optimize|resize|thumbnail) and files that were bulk-accepted on day one (via: analyze). Theviafield (#24) tells you why an entry exists, but not what it is worth: a bulk-accepted file at 3% estimated savings is fine where it is, while one at 85% is the debt worth burning down.What it should do
analyze <dir>table.via, so accepted debt and processed files are distinguishable at a glance.--jsonoutput for scripting.optimize/convert(which re-record automatically) oranalyze --update-baseline.Open questions
glimpse baseline ..., leaving room for future baseline tooling) vs a flag onanalyze?--min-savingsthreshold filter the noise?Came out of the review discussion on #24, alongside #25 (
glimpse init).