Search all three sample catalogues, not one third of them - #17
Merged
Conversation
`examples` read abap2UI5/samples and nothing else, so two thirds of the answer
was invisible to it:
samples 152 apps the PATTERNS - value help, navigation, trees
samples-controls 431 ports the UI5 demo kit, control by control. The answer
to "how do I express sap.m.Wizard", which the
other two rarely have
samples-stack 32 apps the same, but each needs something from the
SYSTEM - OData, RAP, APC, the launchpad. Exactly
what an agent has to know BEFORE proposing one
152 searchable before, 614 now. Each entry names its `repo`, and a new `repo`
filter narrows to one.
This became possible rather than merely desirable: the three catalogues now
render the identical row from the same two lines on the class (`" @summary`,
`" @keywords`), so one parser reads all of them. Before, only samples had a
catalogue in that shape.
A missing checkout is REPORTED, not fatal. One repository absent is not a
reason to refuse the other two - the answer says what was searched and what
was not (`notSearched`), and only all three missing is an error. `examples` is
the one tool here that degrades this way, and AGENTS.md says why: a thinner
answer to "has somebody built this" is worth more than a refusal.
The parser had to change to see the summary at all, and the way it would have
broken is the point. The catalogues grew a second kind of block under the row
title - the sentence, in normal type rather than in `<sub>` - and the old
pattern matched `<br><sub>` blocks only. It would have matched NO ROWS, and
that failure reads as "there are no samples for that" rather than as a parse
error. The blocks are matched as a group and classified afterwards, which is
the same fix the `@docs` links needed a month ago. Tests cover a row with the
sentence, a row without one (the three repositories move at their own pace),
and that the sentence is never mistaken for the search terms.
The three-missing error names every repository, its clone URL and its env var,
and the missing-siblings test asserts all three rather than the one it knew.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
examplesreadabap2UI5/samplesand nothing else, so two thirds of the answer was invisible to it.samplessamples-controlssap.m.Wizard", which the other two rarely havesamples-stack152 searchable before, 614 now. Each entry names its
repo, and a newrepofilter narrows to one.Why this became possible, rather than merely desirable
The three catalogues now render the identical row from the same two lines on the class (
" @summary," @keywords), so one parser reads all of them. Before, onlysampleshad a catalogue in that shape.A missing checkout is reported, not fatal
One repository absent is not a reason to refuse the other two. The answer says what was searched and what was not (
notSearched); only all three missing is an error, and that error names every repository, its clone URL and its env var.examplesis the one tool here that degrades this way andAGENTS.mdsays why: a thinner answer to "has somebody built this" is worth more than a refusal.The parser had to change to see the sentence — and the way it would have broken is the point
The catalogues grew a second kind of block under the row title (the sentence, in normal type rather than in
<sub>), and the old pattern matched<br><sub>blocks only. It would have matched no rows at all, and that failure reads as "there are no samples for that" rather than as a parse error. The blocks are matched as a group and classified afterwards — the same fix the@docslinks needed a month ago.Tests cover a row with the sentence, a row without one (the three repositories move at their own pace), that the sentence is never mistaken for the search terms, and the all-three-missing error. 40 passing.
Generated by Claude Code