i have a computational and memory heavy function that needs to be applied to a different sets of data (read file, run algorithm, collect results). In some cases the algorithm might take to long or use too much memory. My plan right now is too use core async because i assume it will let me isolate each pipeline to its own thread and because it will let me timeout any processes that are running too long (1 minute is probably enough to signify it will never complete). Does this sound reasonable? Is there a way to set limits on how much memory the thread can use up or monitor it? I assume it makes more sense to use the blocking variant via thread rather then a the go block parking version because these last longer, deal with file io, etc...

The JVM has the ability to set the max heap to use for the entire process, but I am not aware of any mechanisms for setting the max heap per thread.

If the work done in a single thread is long enough that starting up a separate JVM for each one is reasonably small time overhead, you could do that for setting the max memory.

is there a testing framework that does clojure and clojurescript, is compatible with the tooling for clojure.test and allows for singling out a test to run very easily ?

or a test runner with an easy ui that does the singling out

@carkh If I want to run a single test, I do it via the REPL with clojure.test/test-vars

Most of the time I run tests via the REPL anyway...

The test runners I've used can narrow to a single namespace but not a single test function I think. Some of them let you put metadata on tests and then filter on that metadata -- so that would let you run a single test I guess.

mhh yes indeed

the metadata trick isn't good for my case i think, tooling won't work with that

What's you use case? Why do you want to run an individual test?

i find that i'm always alternating between running all tests and running a single one

and i like my little desktop notifications

i'm usually not so hot on testing, but i think it very usefull for data structure and algorithmic work

I am pretty big on testing. Of our 87,000 lines of Clojure, about 20k are tests.

While I run full test suites from the command line and occasionally just one namespace, I mostly run tests from my editor via the REPL -- just a hot key to run all tests in the current namespace, or to run the single test under the cursor.

clojurescript is complicating things in this regard

but yes it looks like it works

True, cljs complicates nearly everything 🙂

thanks for the tip

in javascript land there is this neat testing tool where you just change a letter in your test, and now it will only run this one

that would be great with clojurescript's figwhell/shadow-cljs auto-reloading

the eheart of the problem is that when debugging i'm littering the code with logging statements

then i end up with a diarrhea of log entries from all tests

which isn't helpfull at all

so the ideal workflow to me would be : mark the test (via ui or code), add a logging statement, save the file, test is run, check the result and move on with more logging without returning to the test namespace

Is there a way to do a stateful post/prewalk, so that if I detect a particular token, I can do something specific with the next?

I guess I could pull a last-token atom through the walk?