fix(frontend): make a result cell's download work off the canvas

[yangzhang75]

What changes were proposed in this PR?

The download button on a result cell did nothing when pressed from the Form View: the dialog opened, Export closed it, and no file arrived — no request, no error, no message.

An export scopes itself to the operators selected on the canvas unless the caller asked for all of them. That is the right scope for the two callers that read the canvas: the top menu asks for everything, the context menu exports the selection. The third caller is the download button on a result cell, and a cell belongs to one operator, whoever happens to be selected. The same cell is also mounted on the Form View, where the selection holds the step the user is configuring and holds nothing at all until they click one. There the export found an empty scope, so performExport returned at its operatorIds.length === 0 check before sending anything. The dialog's own checks read the same empty selection, so a restricted operator went unreported there too.

A caller that knows which operators it is exporting now says so, and both the dialog and the service prefer that over the selection:

• workflow-result-export.service.ts: a trailing optional requestedOperatorIds on exportWorkflowExecutionResult, threaded to performExport. Precedence is exportAll (whole workflow) > caller-named > canvas selection.
• result-exportation.component.ts: reads operatorIds off the modal data; getOperatorIdsToCheck() returns it when present, and onClickExportResult forwards it.
• result-table-frame.component.ts: the cell's dialog names the operator whose results the frame is showing.

Nothing changes for the menu or the context menu, which name no operators and keep the old scope. Ten lines of logic in three files.

Any related issues, documentation, discussions?

Closes #8538, which has a recording of the before state.

One thing that PR does not change: export-execution-result-enabled still defaults to false in gui.conf, and performExport returns on that switch before it reaches any of this. A deployment with the switch off sees no change from this PR; every deployment that has it on gets a working cell download. Whether that default should flip is a separate question and is noted at the end of the issue.

How was this PR tested?

• Nine new unit tests across the three spec files: the service prefers a named scope over the selection and works with an empty selection, falls back to the selection when none is named, and still lets exportAll win; the dialog names none by default, checks the named operator, and hands it to the export; the frame names its operator, and names none when it has none.
• Mutation check, each restored afterwards: dropping the named scope in the service, in the dialog, or in the frame each turns exactly one named test red.
• Coverage on the three changed files: 100% lines, functions and branches, except one pre-existing unreachable branch elsewhere in result-table-frame.component.ts that this PR does not touch.
• ng test --watch=false: 223 files, 6065 passed, 1 skipped (pre-existing), 0 failed. ng build --configuration=production (AOT): clean. eslint and prettier --check on all six files: clean.
• Exercised in a running instance built from this branch on top of main: a Form View workflow whose result column holds binary files. With the fix, the cell download sends POST /api/executions/result/export/local and returns the bytes (4096 bytes, HDF5 magic 89 48 44 46, byte-exact) while the canvas selection is empty. With the three production hunks reverted on the same stack, the same click sends no request and produces no file within two minutes.

After the fix: downloading a result file from the Form View

Screen.Recording.2026-09-16.at.7.05.42.PM.mov

Was this PR authored or co-authored using generative AI tooling?

Yes. Generated-by: Claude Code (Claude Opus 5, Anthropic). Co-authored with Claude; the author reviewed the change before submission.

🤖 Generated with Claude Code

[github-actions]

Automated Reviewer Suggestions

Based on the git blame history of the changed files, we recommend the following reviewers:

• Contributors with relevant context: @aglinxinyuan, @PG1204, @Yicong-Huang
  You can notify them by mentioning @aglinxinyuan, @PG1204, @Yicong-Huang in a comment.

[github-actions]

Backport auto-label report

This fix: PR was checked against each actively-supported release branch. A release/* label nominates a backport target; the branch's release manager approving this PR is what sends the fix there. The required Backport Approvals check stays red until every label below is approved, so each manager either approves or removes their own label — which is why the labels left on a merged PR are exactly the branches it reached.

Release branch	Analysis
✅ release/v1.3	Already labeled — this fix is queued to backport here. @mengw15 decides: approving sends the fix here, removing this label declines it. The merge waits on one or the other.
✅ release/v1.2	Already labeled — this fix is queued to backport here. @xuang7 decides: approving sends the fix here, removing this label declines it. The merge waits on one or the other.

Auto-label run.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 92.91%. Comparing base (3733b2a) to head (b1510c3).
⚠️ Report is 16 commits behind head on main.

Additional details and impacted files

@@             Coverage Diff              @@
##               main    #8539      +/-   ##
============================================
- Coverage     92.94%   92.91%   -0.04%     
  Complexity     4883     4883              
============================================
  Files          1232     1235       +3     
  Lines         51628    51981     +353     
  Branches       6354     6393      +39     
============================================
+ Hits          47988    48297     +309     
- Misses         2076     2108      +32     
- Partials       1564     1576      +12     

Flag	Coverage Δ		*Carryforward flag
access-control-service	71.78% <ø> (ø)		Carriedforward from 74e6d31
agent-service	99.32% <ø> (ø)		Carriedforward from 74e6d31
amber	88.98% <ø> (ø)		Carriedforward from 74e6d31
computing-unit-managing-service	54.61% <ø> (ø)		Carriedforward from 74e6d31
config-service	87.37% <ø> (ø)		Carriedforward from 74e6d31
file-service	81.53% <ø> (ø)		Carriedforward from 74e6d31
frontend	96.58% <100.00%> (-0.14%)	⬇️
notebook-migration-service	83.73% <ø> (ø)		Carriedforward from 74e6d31
pyamber	98.47% <ø> (ø)		Carriedforward from 74e6d31
workflow-compiling-service	74.09% <ø> (ø)		Carriedforward from 74e6d31

*This pull request uses carry forward flags. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[yangzhang75]

/request-review: @kunwp1

[mengw15]

Dropping the backport labels (they were auto-added from the fix: title, before the rewrite).

The reproduction in #8538 goes through the Form View — its step 2 opens /user/workflow/<id>/form with nothing selected — and the Form View mount of result-table-frame (workflow-form.component.html) came in with #8376 on 2026-09-04, after the v1.3 cut. Neither release branch has it.

[kunwp1]

I left some comments! Overall looks good! Some comments are not directly related to this PR but I want you to fix them because it's a simple fix.

[kunwp1]

LGTM!

[kunwp1]

protected is enough here.