Since my dataflow li­brary Croupi­er is sort-of-­func­tion­al, I need­ed a project where I could ex­er­cise it.

This is im­por­tan­t, be­cause it's how you know if the de­sign of a li­brary is good, ex­ten­si­ble, and so on.

So, I de­cid­ed to write a min­i­mal make-­like-thing.

Well, good news, that was easy!

In about 50 lines of code I could write a thing that will run shell com­mands in a de­pen­den­cy dataflow!

It's called Hacé (don't both­er about how to pro­nounce it, I don't care) which is "im­per­a­tive make, sec­ond per­son sin­gu­lar" in ar­gen­tini­an span­ish, so it's an or­der to make.

I will spend a week or two mak­ing it in­to some­thing semi-use­ful, since it has some ad­van­tages over Make­files, such as re­act­ing to file con­tent and not file date, but its des­tiny is prob­a­bly just to be a test­bed for Croupi­er.

A few re­leas­es of my Crys­tal task/­dataflow li­brary Croupi­er have gone out.

The main top­ic of work has been:

• In­crease code qual­i­ty (I am still learn­ing the lan­guage af­ter al­l)
• Make the API rea­son­able
• Re­move re­stric­tions

On the lat­ter sub­jec­t, Croupi­er will now hap­pi­ly han­dle tasks with ze­ro or many in­put­s, with ze­ro or many out­put­s, or task that share some or all of their out­put­s, even if their in­puts dif­fer.

In all those cas­es it will try to Do The Right Thing, but it is ar­guable whether it does or not.

So, the API and what it can do is chang­ing of­ten. How­ev­er the ex­am­ple in the README on­ly need­ed one change from the first re­lease to now (be­cause it's pret­ty sim­ple)

I know no­body is ev­er go­ing to use it, it's a niche li­brary in a niche lan­guage, but I am hav­ing fun writ­ing it, and the con­cepts are quite wide­ly ap­pli­ca­ble, so it's ed­u­ca­tion­al.