Make the profile the source of truth for author credit - #2093
Conversation
| end | ||
| end | ||
|
|
||
| context 'via search_by_params (the user-facing path)' do |
There was a problem hiding this comment.
🤖 From Claude: Behavior change worth knowing: an author name can no longer be one term of an AND full-text query (search("John technology") used to work). Honoring the credit preference needs per-person branching that a flat SearchCop index cannot express, so person-name search moved entirely to by_credited_person_name, OR-ed in by search_by_params. Single-term author search still works.
| <%= link_to divergence_record_title(record), polymorphic_path(record), | ||
| target: "_blank", rel: "noopener", | ||
| title: "Opens in a new tab", | ||
| class: "text-blue-700 hover:underline" %> |
There was a problem hiding this comment.
🤖 From Claude: Deliberate deviation from the eyebrow convention. These link to 8 different destinations, none of which has a return_to mechanism today (they use hardcoded utility-link rows), so retrofitting all eight was out of scope here — using target="_blank" instead. Happy to do the eyebrow work as a follow-up if you would rather link in-page.
| end | ||
| end | ||
|
|
||
| context 'via search_by_params (the user-facing path)' do |
There was a problem hiding this comment.
🤖 From Claude: Behavior change worth knowing: an author name can no longer be one term of an AND full-text query (search("John technology") used to work). Honoring the credit preference needs per-person branching a flat SearchCop index cannot express, so person-name search moved entirely to by_credited_person_name, OR-ed in by search_by_params. Single-term author search still works.
| <%= link_to divergence_record_title(record), polymorphic_path(record), | ||
| target: "_blank", rel: "noopener", | ||
| title: "Opens in a new tab", | ||
| class: "text-blue-700 hover:underline" %> |
There was a problem hiding this comment.
🤖 From Claude: Deliberate deviation from the eyebrow convention. These rows link to 8 different destinations, none of which has a return_to mechanism today (they use hardcoded utility-link rows), so retrofitting all eight was out of scope — using target="_blank" instead. Happy to do the eyebrow work as a follow-up if you would rather link in-page.
| <%= link_to divergence_record_title(record), polymorphic_path(record), | ||
| target: "_blank", rel: "noopener", | ||
| title: "Opens in a new tab", | ||
| class: "text-blue-700 hover:underline" %> |
There was a problem hiding this comment.
🤖 From Claude: Deliberate deviation from the eyebrow convention. These rows link to 8 different destinations, none of which has a return_to mechanism today (they use hardcoded utility-link rows), so retrofitting all eight was out of scope — using target="_blank" instead. Happy to do the eyebrow work as a follow-up if you would rather link in-page.
bdcfd5d to
c1d2886
Compare
a13d917 to
11fc630
Compare
Name display was asked twice with two value sets: once on the person profile (people index) and once per item (every credit), so the same facilitator could render three different ways with no single place to fix it. Point credits at the profile instead. Anonymity stays a per-item latch — a person may want four stories credited and the fifth not, and nothing should be able to de-anonymize an item that was submitted anonymously. Co-Authored-By: Claude <noreply@anthropic.com>
Strip the per-item credit select from all 8 forms and show a warning only where the stored snapshot disagrees with the profile. The stored column stays as the consent record, editable in one place. Search now honors the preference too: an anonymous credit matches nothing, first_name_only isn't findable by last name, and first_name_last_initial matches only the initial. Dropped the person_first/person_last SearchCop attributes that bypassed this. Co-Authored-By: Claude <noreply@anthropic.com>
Main introduced shared/_badge while this branch was in flight; the reconciled pill was hand-rolled. Co-Authored-By: Claude <noreply@anthropic.com>
Sort each person's suggested preference into the worklist and name it inline (e.g. "Suggested: First name only — most restrictive across Maria Johnson's content") so the hint is specific to the person rather than a generic line. Add a person name-search picker to the filters and order groups by first then last name. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Keeps the admin worklist open while checking a profile, matching the content links that already open in a new tab. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Clearing a single column is what makes that column safe to drop, so each one gets its own subsection and its own all-clear note naming the field. Guess who each free-text name refers to so an admin confirms rather than looks every one up, resolved in one query for the page. Co-Authored-By: Claude <noreply@anthropic.com>
Every fix on the author credit divergences page (Save stored consent, Apply to profile, Credit to a person) did a full-page redirect. Respond with a Turbo Stream that re-renders just the results frame and flash, so the worklist updates in place with no page flip. Non-Turbo requests still redirect. Also widen the remote-select left padding so the search icon never overlaps the placeholder or selected value. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- author_credit returns "Anonymous" for a stored-anonymous item even when it carries a legacy free-text name — that name belongs to no profile, so nothing else would suppress it. - Add a credited_openly scope so a NULL snapshot (which means "follow the profile") stays visible; a bare where.not would drop it, since NULL never compares unequal. Use it for a person's public authored content. - Build the divergences result only on the frame request, so the full page load stays cheap. - Under an active filter, an empty section is the filter's doing, not a milestone, so section-clear withholds the congratulations and cleanup note. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The divergences page couldn't finish its own job. An item whose stored consent was "anonymous" offered no way back to "follow the profile", so the only exit was picking a name format — re-recording a consent nobody gave. Per-item anonymity is the legacy state this page exists to drain, and the profile is the source of truth, so clearing it is now allowed on every row. The author picker fell back to the creator, so a workshop credited to the legacy name "Lisa Cohen" showed "Umberto User" in its own edit form — and the next save would have written that creator into author_id, destroying the legacy credit. The picker now reflects only the record's own author (still defaulting to the creator on new records) and names the legacy credit standing in when there isn't one. Sorting COALESCEd author, creator, legacy while display used author, legacy, creator, so legacy-credited rows sorted under a name they never show. A note under each author picker flags when the credited person's profile suppresses credits, since nothing on the record itself reveals that. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
…the creator The column reads better as the state it is than as a predicate on a collection, and it now matches the "Anonymous contributions" label already used in the profile form and the factory trait. The migration is unmerged, so it changes in place rather than stacking a rename on top of itself. A legacy free-text name follows nobody's profile, so a legacy-credited row now reports no governing person. It was being grouped on the divergences page under whoever happened to enter it, asserting that person's profile had drifted from a consent snapshot that never described them. Those rows have to be matched to a real person by hand against the legacy text — the creator is not a fallback for them. Community news author is optional on the model, so the form offers a blank option rather than forcing whoever edits one of the legacy authorless rows to credit somebody before they can save anything else. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The table had drifted well past this branch's additions. Co-Authored-By: Claude <noreply@anthropic.com>
11fc630 to
e4c6167
Compare
Co-Authored-By: Claude <noreply@anthropic.com>
The workshop show page rendered a Resource's author association directly, which has no to_s — so a public page printed nothing where a credit belonged, honoring neither the name preference nor anonymity. The workshop idea page used User#full_name, the real name, where its sibling idea pages use User#name. Also drops the obsolete brakeman ignore entry for callouts_controller; the warning it silenced no longer fires. Co-Authored-By: Claude <noreply@anthropic.com>
… fallback Two paths still routed a legacy free-text credit through its creator. The consent snapshot was taken from the creator's profile on create, so a legacy row was born diverged from a profile that never described it — the exact rows that shouldn't be resolved that way. And credited-name search matched the creator, so a workshop displaying "Lisa Cohen" came back under the name of whoever typed it in, which is the mismatch between search and display the rest of this work removes. The portal sits behind a login, so an unattributed credit says "AWBW Facilitator" rather than "AWBW Staff" or "Anonymous". "Anonymous" in particular read as a deliberate privacy choice when it only ever meant nobody was recorded. That makes every per-model override redundant, so they come out. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Entering a record isn't claiming it, so a model that can name an author now credits only that author. Where none is named the credit reads "AWBW Facilitator" rather than the name of whoever typed it in — and it's no longer findable or sortable under that person either, which was the last way the creator leaked into a credit they had nothing to do with. The idea models keep their creator credit: they have no author_id at all, so it's the only attribution they can carry. A suppressed credit reads "AWBW Facilitator" too, from either anonymity path. Behind a login, "Anonymous" invited the reader to wonder who was being hidden; the generic credit just says the org stands behind it. The suppression is unchanged — the name is still withheld and still never links. Section 3 of the divergences page follows: those records show the generic credit now, so the creator appears there as a suggested author to confirm, not as the profile that formats them. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
🤖 suggested review level: 5 Inspect 🔬 changes how every author credit renders app-wide, narrows search results, and adds an admin reconciliation page
Name display was asked in two unrelated places with two different value sets —
people.display_name_preference(people index) and<item>.author_credit_preference(every credit, on 8 forms).AuthorCreditabledeliberately calledperson.full_namerather thanperson.name, so the per-item answer won and the profile was ignored. One facilitator could render three different ways with no single place to fix it.Now: the profile formats every credit. The 8 per-item selects come off. The stored column is retained as the consent record, human-editable only on a new admin reconciliation page.
Two axes on the profile
display_name_preference(4 values)Person#nameand author creditsanonymous_contributionsanonymous_contributionsalready existed on main (added by20260810225127) but was inert: nothing read it. This PR makes it actually govern credits, soStoryImporterflagging an anonymous import now genuinely suppresses that credit.The one asymmetry
A stored
author_credit_preferenceno longer drives display — exceptanonymous, which is always honored. Anonymity is true if the profile says so or the item says so; neither can strip it from the other. A suppressed credit renders the generic "AWBW Facilitator", not the word "Anonymous" — behind a login that invited the reader to wonder who was being hidden, where the generic credit just says the org stands behind it. The suppression itself is unchanged: the name is withheld and never links. Anonymity is inherently per-item (a person may want four stories credited and the fifth not), admins need a per-item lever, and there is no backfill — so without this, every item currently storinganonymouswould start rendering its author's real name on deploy.Search honors the preference
Anonymous → matches nothing.
first_name_only→ not findable by last name.first_name_last_initial→ matches the initial only. This also meant dropping theperson_first/person_lastSearchCop attributes onStory/CommunityNews, which bypassed the gate. Tradeoff: an author name can no longer be one term of an AND full-text query — honoring the preference needs per-person branching a flat index can't express.Reconciliation page (
/author_credit_divergences, linked from admin home)Four sections, in the order to work them. Each names the code it retires once cleared, so the legacy credit paths can actually be deleted:
workshops.full_name,resources.legacy_author_name) so each field can be retired on its own, with a suggested person matched from the free-text nameauthor_id, so the credit falls back to the creating user's person (idea models excluded — that's their only credit path)Sections 2–4 all resolve by assigning an
author_id, the only credit path that follows a profile, links to it, and lists the record there. Filterable by person (remote search), content type, stored preference, and reconciled state.Only the author is credited
Entering a record isn't claiming it, so a model that can name an author credits only that author — no creator fallback in display, search, or sort. Where none is named the credit reads "AWBW Facilitator". The idea models are the exception: no
author_idcolumn at all, so their creator is the only attribution they can carry, andcredits_creator?draws that line in one place. Section 3 of the reconciliation page follows — those records already show the generic credit, so the creator is offered there as a suggested author to confirm, not as the profile that formats them.Legacy credits stay clear of whoever entered them
A legacy free-text name outranks the creator, so those rows have no governing person. They take no consent snapshot on create (they were being born diverged from a profile that never described them), they aren't grouped under their creator in section 1, and credited-name search no longer returns them under the creator's name — a workshop displaying "Lisa Cohen" was coming back under the name of whoever typed it in. They resolve only by matching a real person against the legacy text. The author pickers follow the same rule: they show the record's own
author(defaulting to the creator on new records only), so saving an untouched form can't silently promote the creator over a legacy credit.One fallback label
An unattributed or suppressed credit reads "AWBW Facilitator" everywhere. The portal is behind a login, and the old three-way split ("AWBW Staff" on community news/resources, "AWBW Facilitator" elsewhere, "Anonymous" on the idea models) made "Anonymous" read as a deliberate privacy choice when it only ever meant nobody was recorded. All five per-model overrides are gone.
Also swept
Four places where a real name escaped past an anonymized credit: the variation-ideas index link, the story form's subtitle,
Workshop#author_name, and two deadResourceDecoratormethods. Plus two phantom strong params (profile_show_name_preference,promoted_to_story) referencing columns that don't exist.Verified
Full suite green (6898 examples, 0 failures, 17 pre-existing pendings) via
ai/test_extra. Rubocop clean on 1511 files. Driven in the running app: format changes propagate to the index and every credit at once, both anonymity paths suppress credit and link, search gating confirmed against real data, all four reconciliation sections populate and resolve.🤖 Generated with Claude Code