EE-4048: add attributeFriendlyName to every application template (819 of 828)

[nsaraceno-sa]

Draft on purpose. Merging this alone ships nothing — see This does not deploy itself below. It is also an alternative to swap-monorepo#209, which is already open and does the same job in one place. Opened so the diff can be looked at before choosing.

What

SecureAuthCorp/IdP#426 (EE-4048) made FederationAttribute.friendlyName persist. Until then ExtractAttributes copied five of the six fields and left FriendlyName null, so the realm writer's if (attr.FriendlyName != null) guard skipped SAMLAttr{n}FriendlyName. The field is writable now; the SWAP attribute editor has nowhere to type it.

Two additions per template:

// entries
"attributeFriendlyName": {
  "label": "Friendly Name",
  "controlType": "input",
  "dataModelName": "friendlyName"
}

// statics.attributeCallout.after
[ "attributeInfo", "attributeFriendlyName", "attributeNamespace", ... ]

Placed directly after the name/property grid rather than appended after attributeFilterGroup: it decorates the attribute name, and the realm writes SAMLAttr{n}FriendlyName beside SAMLAttr{n}Name.

Coverage

819 of 828 templates. Diff is +4,922 / −15, about six lines per file.

The nine skipped are top-level JSON arrays rather than objects — ApexAnalytix, Firmplay, Miro, SageX3, VictorOps, Wombat, Wrike, ZscalerInternetAccess, ZscalerPrivateAccess. template-compiler already drops them, because template.data.templateName is undefined for an array, so they are absent from the compiled set and inert in production today. They want their own fix, not a workaround here.

Also already inert, for the same kind of reason, and untouched here: CrashPlan, MobileIron and PartnerCentral have no icon (the compiler skips them), and Atipica.Json's capital J means it compiles to Atipica.Json.json.

Reviewing the diff

Generated by script, preserving each file's own formatting — 2-space indentation and its trailing-newline state.

This matters more than it sounds. Re-serialising at the compiler's own 4-space indent instead rewrites all 38 MB: 866,000 added and 861,000 removed lines, median 2,110 per file, versus ~4,900 total here. If you regenerate this branch, keep the 2-space indent.

Salesforce.json carries twelve extra lines — it uses Allman brace placement ("notify": then { on the next line) that JSON.stringify normalises. Cosmetic, and the only file with churn beyond the six added lines.

Ellucian_Colleague.json shows a replaced line rather than an added one: its after is a single-element array, so the last entry gains a comma.

Verified equivalent to the runtime injection

swap-monorepo#209 injects this same entry into each template as dagobah fetches it. Running that injectEntry over all 816 compiled templates and comparing entries.attributeFriendlyName plus statics.attributeCallout.after against these patched sources:

compared: 816   identical: 816   differ: 0

So this is a true drop-in. If it lands, #209's augmentTemplate.js and its yaml fragment can be deleted in the same change with no behavioural difference — the swap-monorepo side becomes a deletion.

This does not deploy itself

template-compiler/compile.js is a pure download-and-write (JSON.stringify(template.data), no transform), and CI does not run it. .github/workflows/template-compiler.yml dropped the build step deliberately:

Build stage for "npm install" and "node compile prod" commands where not included in this workflow since the templates are deprecated and wont change.

It only does aws s3 cp ./dist/ from the committed template-compiler/dist in swap-monorepo. So shipping this needs, in order:

1. this PR merged;
2. node compile prod run against it, and the ~816 regenerated files committed to swap-monorepo's template-compiler/dist — that diff is clean, six lines per file (simulated);
3. develop/master → dev bucket, live → prod.

Step 2 is in the directory swap-monorepo's CLAUDE.md marks legacy and do-not-touch, and this change falsifies the "templates … wont change" assumption the workflow was simplified around. Worth deciding whether that step gets restored to CI before relying on this route again.

Trade-off against #209

This way: the field lives in the templates, so any consumer of them gets it, and no runtime code is needed. Costs a three-surface deploy, and every new template added here has to remember the field.

#209: one file in one sub-app, covers every template present and future, ships on the normal SWAP release. Costs a small amount of runtime code that has to be maintained.

Refs: https://secureauth.atlassian.net/browse/EE-4048
Refs: https://github.com/SecureAuthCorp/swap-monorepo/pull/209

🤖 Generated with Claude Code