Resync prob_dist.md to source ebff90fd (recovers four sync failures) - #300
Conversation
The sync of this lecture failed four times in a row — lecture-python-intro #812, #818, #820 and #822, one per source PR, all structural-parity failures, all filed 2026-08-03 and never answered. The target has been stuck at the 2026-07-18 bootstrap sha ever since, missing three of the four source changes: the Observed Distributions split (#811), the figure captions (#816) and the sequence linking (#819). Only the data-lectures reads (#821) reached it, by hand in #298. Recovered with `translate forward -f` at engine v0.26.0 rather than \translate-resync, per the 2026-08-20 ruling on QuantEcon/action-translation#276. Verified against source: 55/55 directives, 49/49 code cells, 14/14 headings, 0 python syntax errors across all 49 cells (matching source), and the i18n font block preserved. Two things worth recording. All 19 figure captions landed and all 19 are localised, none byte-identical to source — so the caption class tracked in QuantEcon/action-translation#255 does not reproduce on the forward path, only on sync and init. And the hand-applied data-lectures reads from #298 survived the whole-file regeneration intact. The five axis labels that match source are CDF, PDF, PMF, S and x — acronyms and symbols, correctly left alone. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
✅ Deploy Preview for playful-platypus-17d3bb ready!
To edit notification comments on pull requests, go to your Netlify project configuration. |
There was a problem hiding this comment.
Pull request overview
This pull request regenerates lectures/prob_dist.md to resync the translated lecture with the upstream source commit ebff90fd, and updates the corresponding translation state metadata to reflect the new sync point and tool version.
Changes:
- Restructures/expands the probability distributions lecture content (e.g., moments/quantiles discussion) and updates section organization.
- Adds/updates
mystnbfigure captions and labels across the lecture’s plots. - Updates
.translate/state/prob_dist.md.ymlwith the newsource-sha,synced-at, andtool-version.
Reviewed changes
Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.
| File | Description |
|---|---|
lectures/prob_dist.md |
Whole-file resync of the translated lecture content and figure metadata to match upstream structure/captions. |
.translate/state/prob_dist.md.yml |
Advances the recorded source SHA/date and updates the translation tool version metadata. |
💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.
| @@ -430,6 +531,12 @@ u.mean(), u.var() | |||
| 这里是部分PMF： | |||
Correction to the "#255 finding" aboveI claimed the caption defect looks path-specific — that The accurate picture, across the four lectures regenerated on 2026-08-20 at engine v0.26.0:
So Nothing in this PR changes — its 19 captions came through localised from the engine and were not touched by hand. But the generalisation was wrong, and the sharper claim is that caption localisation is unreliable rather than absent, which is a different thing to design a check against: a deterministic assertion is still needed, because you cannot tell from a given run whether it happened to work. Recorded on action-translation#255. |
Brings
prob_dist.mdup to sourceebff90fd. It has been stuck at the 2026-07-18 bootstrap sha since the sync started failing on it.Why it was behind
The sync failed on this lecture four times in a row, once per source PR, and nobody ever answered any of them:
All four are structural-parity failures. Only the last one's content reached this edition, applied by hand in #298 — which is why the state file looked merely stale while the lecture was genuinely 216 lines short and carrying 3 captions against source's 19.
How it was delivered
translate forward -f prob_dist.mdat engine v0.26.0, not\translate-resync. Per the 2026-08-20 ruling on action-translation#276 the CLI is the documented recovery route: it reads source off the local filesystem, so it cannot carry a stale merge-time snapshot, and its parity check is anchored to currentmain.Verification
A finding worth carrying to #255
I expected this recovery to walk straight into the untranslated-caption class, since it delivers 16 new captions. It did not: all 19 came through localised. The same engine at the same version left all four of
mobility.md's captions in English when run throughinit(#299), and the class is what led #296 to be closed unmerged on the sync path.So the caption defect looks path-specific —Retracted — see the comment below. That generalisation was wrong and rested partly on a hand-fix rather than engine output. Measured across all four lectures regenerated that day:forwardhandles captions,initandsyncdo not.prob_dist19/19 localised,hoist_failure7/7,learning_approximation4/4 — butmobility0/4, same session and engine version. Soinitis inconsistent run-to-run, not uniformly broken. The sharper claim, now recorded on action-translation#255, is that caption localisation is unreliable rather than absent — a harder defect to design against, because each individual run looks fine.Review note
This is a whole-file regeneration, so the prose is re-translated rather than patched. The Observed Distributions split means a large part of the lecture is genuinely new here, and the 19 captions were written by the model — both worth a native read.