Repository navigation
Let students sort search results by lowest rated - #70
Merged
Merged
Conversation
Lowest rated uses the same score as Highest rated, reversed: each entity counts as if it had two extra 3-star reviews, so one bad review does not put it on top. Ties go to the entity with more reviews, and entities with no reviews still go last. Also quote file paths in the pre-commit hook, so it works in a checkout whose path has spaces.
coldfinity
pushed a commit
to coldfinity/monmap
that referenced
this pull request
Nov 14, 2026
Reported on A2000 in 2022: selecting a major offered "Extended major
2" and "Extended major 3". The repeating slots were working as
designed; the labels were wrong, and had been in every year except
2026.
A2000 major extended_major
2022 1 27
2025 1 29
2026 29 1 <- same AoS codes
This is feedback monashcoding#66/monashcoding#74/monashcoding#70 ("Arts majors and extended majors are
swapped; psychology is the only major"). The triage recorded it as
closed on the strength of 2026 alone — the same single-year blind spot
the cross-year pass caught in the plan. Two defects in
classifyAosRelationship, both visible in the real labels:
A container can list both kinds. 2022-2025 A2000 uses "Part A. Arts
listed majors and extended major" for the whole list, which 2026
splits into "Part A. Major studies" plus one separate extended major.
Testing for "extended major" first labelled all 27-29 extended,
inverting the course. When both appear, the plural listing is the
container's subject and the extended one is the aside, so the bulk are
majors — which is what 2026 resolves the same codes to.
Whitespace varies. "APAC - Psychology extended major" is verbatim
from the corpus, double spaces and all, and an exact "extended major"
test missed it — so a genuine extended major became the only thing
A2000 called a major. That is the "psychology is the only major" half
of the report, and it was wrong in 2026 too.
Attribution was measured rather than assumed: running both the old and
new classifier over every stored label in all seven years changes 353
rows across exactly 3 distinct labels, all of them these two cases. A
full --force backfill also carries pending drift in the older years
from earlier extractor changes (about 50 rows: specialisation ->
elective, other -> major), which the same measurement separated out and
which aligning those years with the current extractor was going to
apply anyway.
Backfilled all seven years. A2000 is now consistent everywhere: 26-31
majors and 1-2 extended majors, no inversion. The only resolver
snapshot drift is D3002 going 29/1 to 28/2 as psychology moves, with
aosGroups and aosGroupOptions unchanged. A2000 is added to the harness
at both 2022 and 2026 so a single-year fix can never look complete
again.
Verified: typecheck clean, 183 webapp tests, 45 db, 27 ingest (6 new,
using the real corpus labels), resolver snapshot OK at 17 cases,
corpus baseline OK.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
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.
Lowest rated uses the same score as Highest rated, reversed: each entity counts as if it had two extra 3-star reviews, so one bad review does not put it on top. Ties go to the entity with more reviews, and entities with no reviews still go last.
Also quote file paths in the pre-commit hook, so it works in a checkout whose path has spaces.