A tiny C program used to periodically execute a command.

Usage: watch [options] <cmd>
Options:
  -q, --quiet           only output stderr
  -x, --halt            halt on failure
  -i, --interval <n>    interval in seconds or ms defaulting to 1
  -c, --clear           clear the screen between iterations
  -v, --version         output version number
  -h, --help            output this help information

$ make install

Or in your local bin (~/bin)

$ PREFIX=~ make install

This project is very similar to original watch(1) implemented in 1991, differences include:

• ansi escape sequences (colors etc)
• terminal is not cleared (unless --clear is provided)
• lower default interval of 1s
• millisecond interval resolution

This version of watch(1) support millisecond resolution with the ms suffix:

$ watch -i 300ms echo hey

whereas 300 would be seconds:

$ watch -i 300 echo hey

Watch is pretty handy, here are a few use-cases:

Ad-hoc mtime watchers are annoying to construct, and have relatively no purpose when you can simply execute your tests at a regular interval. For example run watch(1) as a job, running tests each second (or a second after the program exits):

$ watch make test &
[1] 3794
✔ bifs.components
✔ bifs.dark
✔ bifs.darken
✔ bifs.image-size
...

Your tests will happily chug away, when you want to stop watch simply foreground the job and ^C:

$ fg

Need to build CSS or JavaScript dependencies? use a Makefile. With the large quantity of copy-cats (Rake,Jake,Sake,Cake...) people seem to be forgetting that Make is awesome, if you take a little bit of time to learn it you'll love it (or at least most of it). Make will utilize mtime and only build what's necessary, this is great.

Let's say we had some Jade templates, even some nested in sub-directories, we could list them in a Makefile quite easily.

Below JADE is a list constructed by the shell command find templates -name "*.jade", which is usually a lot easier to manage than listing these files manually, which is also valid, and sometimes important of ordering is relevant. Following that we have HTML which simply substitutes ".jade" with ".html", giving us our HTML targets.

JADE = $(shell find templates -name "*.jade")
HTML = $(JADE:.jade=.html)

Our first target is all, becoming the default target for make. On the right-hand side of this we specify the dependencies, which in this case is a list of all of our HTML files, not yet built. Make will see this and execute the %.html targets, which allows use to use the jade(1) executable to translate the dependency on the right of :, to the target on the left.

JADE = $(shell find templates -name "*.jade")
HTML = $(JADE:.jade=.html)
all: $(HTML)
%.html: %.jade
	jade < $< > $@

Now we can build all of these files with a single command make:

$ make
jade < templates/bar.jade > templates/bar.html
jade < templates/baz/raz.jade > templates/baz/raz.html
jade < templates/foo.jade > templates/foo.html

We can also add a clean pseudo-target to remove the compiled files with make clean. Here it's listed to the right of .PHONY:, telling make that it does not expect a file named ./clean on the fs, so it wont compare mtimes etc. Make is smart about re-executing these actions, if you make again you'll notice that since none of the dependencies have changed it'll simply tell you "make: Nothing to be done for `all'.".

JADE = $(shell find templates -name "*.jade")
HTML = $(JADE:.jade=.html)
all: $(HTML)
%.html: %.jade
	jade < $< > $@
clean:
	rm -f $(HTML)
.PHONY: clean

The one missing component is periodical action, which is where watch(1) or similar utilities come in, this functionality coupled with Make as a build system creates a powerful duo.