hoboneer / redo Goto Github PK

In line with the distinguishing of normal, phony, and multi-output targets proposed by #3, these distinctions could be used for multi-output targets.

Maybe redo-addtarget could register the target with the central database, and give the caller the name of the tempfile to write to. The use of this command would mark the current target as virtual.

A central .redo directory (a la Git) simplifies this. So, where should it be stored? Should it depend on cwd of the toplevel redo call?

Since we already know the dependency graph, having automatic cleaning is convenient and safe.

By default, redo-clean TARGET should remove TARGET only. A flag -r or -R would recursively delete TARGET, its dependencies, their dependencies, and so on until we clean the leaf nodes in the dependency graph.

From what i've seen in the wild, targets that, in make, would be marked .PHONY aren't anything special in the eyes of redo.

They're just run with redo, while their dependencies would be run with redo-ifchange.

What happens if a file with the same name exists? If used with redo, there's no problem, but if used with redo-ifchange, nothing is run.

Distinguishing real, phony, and "virtual" (multi-output) targets might help:

1. users unfamiliar with the repo, since a command like redo-targets -p (p for phony) could be run; and
2. other programs, which could use that information for other purposes (such as autodocs, or linting the project's .do files?)

As in make, the use of shell allows easy tracing (and thus transparency) with set -xv.

This could be generalised to any executable .do files, capturing stderr as the tracing output.