I love Makefiles and use them in almost all my projects, but I think the main benefits of them is walking DAGs of dependencies, which the article doesn't really mention.
# final-file depends upon the state of intermediate-file
final-file: intermediate-file
tool-b intermediate-file >final-file
# intermediate file depends on source-file
intermediate-file: source-file
tool-a source-file >intermediate-file
Then the command 'make final-file' will run tool-b to generate the intermediate file, and knows that it first needs to run tool-a to generate the intermediate-file, and (the clever bit) it only runs each step if the output of the step is older than any of its inputs.
This means you eliminate needless typing (from running a tree of pre-requisite steps), and you eliminate needless (and possibly slow) rebuilding of files which don't need it. This is the core value of make, and if you can't take advantage of this, then there isn't much value over a set of script files.
I'd classify my current employer (Canonical) as high on the technically savvy scale, and my team uses a Makefile or two on every project. Even in cases where we can't make use of the above dependency analysis, it's nice for us to standardize on one approach to controlling how to build/test/etc projects.
I love Makefiles and use them in almost all my projects, but I think the main benefits of them is walking DAGs of dependencies, which the article doesn't really mention.
For example, if your toolchain generates files:
Then you can define some Makefile targets:
Then the command 'make final-file' will run tool-b to generate the intermediate file, and knows that it first needs to run tool-a to generate the intermediate-file, and (the clever bit) it only runs each step if the output of the step is older than any of its inputs.
This means you eliminate needless typing (from running a tree of pre-requisite steps), and you eliminate needless (and possibly slow) rebuilding of files which don't need it. This is the core value of make, and if you can't take advantage of this, then there isn't much value over a set of script files.
I'd classify my current employer (Canonical) as high on the technically savvy scale, and my team uses a Makefile or two on every project. Even in cases where we can't make use of the above dependency analysis, it's nice for us to standardize on one approach to controlling how to build/test/etc projects.
Very interesting, thanks.
I added some missing "dependencies" (the parts after the colon) in my Makefile example.