: about.[ Home | Source | Mailing list | Documentation ]
There are no shortage of build systems, what justifies the existence of yet another one?
Mem is a completely different approach to build systems, one base on function memoization. Most build systems are based on building up some static graph of dependencies and targets, traversing that graph while minimally rebuilding, propagating changed objects across the graph for calculating further steps; usually with very special languages and sophisticated layers for automatically building the graph, easing the job of the end user in describing their build.
Practically, mem is different in that:
the build is simply a script where certain functions are memoized (cached)
these memoized functions are only rebuilt if one of their inputs changes
The first part means that you can script your build however is most natural to solve your specific build requirements, and mem sits in the background and caches and avoids duplicate evaluations (eg needless rebuilding).
The second part means exactly what it says. This can be startling at first. For instance, suppose you had a .c file, and a build step to compile it into an object and a build step to link that into a program. If you make a whitespace change to that .c file, such as changing a comment; the resulting object will likely be identical (sha1) as the object before this change was made. Given that mem only builds when things are changed, no relinking will be done as the object to link will be unchanged!
Mem requires Python 2.5 or greater.