Re: Hot to write pattern rules for suffix deletion

On Thu, Jan 4, 2024 at 8:51 PM Stephen Touset <[email protected]> wrote:

> I have a set of files in a project that need to be preprocessed through
> `envsubst` to render their output. The targets of these templated files
> are a mix of different file types, so they aren't rendered to files with
> a consistent extension. Further, not all files with those target
> extensions need to be rendered from templates. For example:
>
>     a.yaml.envsubst    // a.yaml should be rendered from this
>     b.json.envsubst    // b.json should be rendered from this
>     c.json             // should not pass through a rendering step
>
> I would like to express this in a make rule. Instinctually, I want to
> reach for something like
>
>     %: %.envsubst
>         envsubst < $< > $@
>
> Obviously this rule isn't correct.


Can you explain why you don't believe that to be correct?  Given that a
pattern rule only applies when the prerequisites either exist or make knows
how to create them, why wouldn't that work?



> But I'm a bit at a loss for how to
> express what I want through make rules. One approach that I considered
> was using `$(shell find ...)` to get a list of these templates and
> generate rules for each one. Unfortunately in my case, that would be
> quite difficult because one file might go through *multiple*
> preprocessing steps:
>
>     d.json.envsubst.gz // d.json should be decompressed then rendered
>

That could be handled in either of two ways.  If .gz only needs to be
handled in the exact .envsubst.gz combo, then simply having a second
pattern for those that does the right thing would work:

%: %.envsubst.gz
        gzcat $< | envsubst > $@

Alternatively, if you use .gz in combo with all sorts of suffixes AND
decompressing to a file and then operating on the file isn't unreasonable
or a pessimization that you want to avoid, then you can just have a general
rule for .gz files that make can figure out how to chain with the first
above:

%: %.envsubst
        envsubst < $< > $@
%: %.gz
        gzcat < $< > $@


So now I'd need to have some overcomplicated logic to write out the
> resulting rules:
>
>     d.json: d.json.envsubst
>         ...
>
>     d.json.envsubst: d.json.envsubst.gz
>         ...
>

Why?  I don't see why you would need to do that.


Philip Guenther