build_pipeline.md

Suppose I have a bunch of CoffeeScript files:

> ls src
foo.coffee    bar.coffee    baz.coffee

I want to produce the following for deploying:

• a concatenated, gzip'ed, hashed file suitable for serving up in production
• a source map
• documentation
• unit test coverage

Thus, I need a build pipeline like so:

foo.coffee  |
bar.coffee  |>--->all.coffee--->all.js--->all.min.js--->all.min.gz--->all.min.9845iegiwfhfgaksjfsad.gz
baz.coffee  |               \--->all.sourcemap
                             \---->all/*.html
                              \
foo_spec.coffee  |             \---v
bar_spec.coffee  |>------------------>coverage.json
baz_spec.coffee  |

If this is modeled as a build pipeline, i.e what make from 1970 was created to do, the build will know that if I change foo_spec.coffee and wish to re-calculate my test coverage, I do not need to concatenate my source files, as they have not changed.

This is obviously simplified, but in a real app, there would a lot of build artifacts that depend on one another, and you don't wanto rebuild the world everytime you run the build or everytime you change one file.