target name with a colon not showing inputs correctly

[lourd]

Current Behavior

I have a project in my monorepo that has a target named test:integration. I was trying to debug some caching misses and was confused that it was showing as having no inputs. When I changed the named to test-integration, dropping the :, the inputs showed up as expected. Nothing else changed in my configuration in between these two screenshots

Expected Behavior

Targets with a colon in the name should show their inputs like any other target name

GitHub Repo

No response

Steps to Reproduce

1. Create a target with a colon in the name
2. Open it in nx graph
3. See that it doesn't show any inputs

Nx Report

Node           : 22.14.0
OS             : darwin-arm64
Native Target  : aarch64-macos
npm            : 10.9.2

nx             : 22.1.3
@nx/js         : 22.1.3
@nx/workspace  : 22.1.3
@nx/devkit     : 22.1.3
typescript     : 5.9.2

Failure Logs

Package Manager Version

pnpm 10.11.0

Operating System

• macOS
• Linux
• Windows
• Other (Please specify)

Additional Information

No response

[FrozenPandaz]

Thanks for the clear repro! Confirmed as a real bug — root cause is in nx graph's server-side task-input resolution, not the UI rendering logic.

Root cause

packages/nx/src/command-line/graph/graph.ts:1296 naively splits the taskId on ::

const [projectName, targetName, configuration] = taskId.split(':');

For a taskId like myproj:test:integration, this produces targetName = "test" and configuration = "integration" — but the real target is test:integration with no configuration. Downstream, createTaskGraphForTargetsAndProjects(['test'], ['myproj'], 'integration') finds no target named test on the project (the filter at graph.ts:1224 drops it), plans[taskId] is undefined, and getExpandedTaskInputs returns an empty object — so the UI has nothing to display.

Fix

Use the existing colon-aware helpers at packages/nx/src/utils/split-target.ts instead of a raw .split(':'). The groupJointSegments logic there (L277) already handles multi-colon target names by consulting the project graph's known target names. Something like:

const [projectName, targetName, configuration] = splitTargetFromNodes(
  taskId,
  Object.fromEntries(
    currentProjectGraphClientResponse.projects.map((p) => [p.name, p as any])
  )
);

(or equivalent — whatever lets splitTargetFromNodes see the project nodes keyed by name).

Testing

A unit test that calls getExpandedTaskInputs("myproj:test:integration") against a project whose targets contains "test:integration": { inputs: [...] } and asserts the returned object is non-empty would lock this in.

Community contributions welcome — this is a well-scoped fix!