msackman / argot Goto Github PK

argot.go:28: Achieved Steps: [0x577810 0x577990 0x577c20 0x577dd0 0x578090];

As much as I enjoy memory addresses in my debug logs...

So I know that a certain project uses Go() error for other purposes, but it really looks to me like that's hindering the ability to reuse steps and compose them because e.g. the reporting on the last failed step will only ever report the top-level steps that failed, not the child steps that may have been the original reason.

Of course, wrapping the errors with some more helper code could work, but I wanted to put this up for discussion in case I might have just missed an obvious idea here.

E.g. I have a step NewGitRepository(<my repo>, <my commit>) that's composed of MakeATemporaryDirectory, CloneInto(<my repo>, <the temporary directory>), and Checkout(<my commit>) - if either of the substeps fail, the whole step fails, as intended, but the error message is somewhat disconnected from where the error occured.

AFAIR the channel code wasn't really working out, but for reuse purposes some better form of composition would be nice - either the step itself does the reporting and knows how to untangle its substeps, or the step will always "flat map" all it's substeps such that at the top only a simple (not nested) list of steps will be executed.

Thoughts?