calpaterson

calpaterson, 8 days ago

@hl allow me to share my process with you:

1. i write it, and edit and rewrite over and over. after many rounds of edits i get to a final draft

2. then i have my wife proof-read it. she is a former book publisher and very very good at finding typos and mistakes. we take them all out

3. i do another round of edits on my own, putting lots of new mistakes in

4. publish

calpaterson, 12 days ago

@lewiscowles1986 I have never tried mutation testing, but I am interesting in giving it a go.

Property based testing I do use in certain places but more often I am doing example-based testing with parameterised examples. My reservation is that every commercial project I've seen it used it it's been too slow

calpaterson, 12 days ago

@lewiscowles1986 i think broadly speaking 30s is a good target. i currently get 9s without using parallelism (but parallelism does work)

mutmut seems pretty interesting

re:sqla 2.0 - close to ready. most (all?) of the future=True switches are on. Just other stuff to do

calpaterson, 11 days ago

@jamescooke I had a quick look at mutmut yesterday. It got working with no config which was impressive

I ran it for about an hour or so while I did nursery run.

It ran about 500 mutations in that time. 50% passed and 50% failed. of the failures:

1. a small minority were code i prefer not to test, like the exact contents of log messages

2. all the rest were dead code that should be removed

so a useful exercise, and i'll carry on with it, but no bugs found yet

CC @lewiscowles1986

calpaterson, 11 days ago

@jamescooke @lewiscowles1986 you can annotate certain lines to be immune from mutation which looks essential for non-trivial codebases

calpaterson, 11 days ago

@lewiscowles1986 @jamescooke here's an example of a mutation that i didn't care about

it's possible to add an assertion to cover this, but i am happy that it's typechecked + runs through

i don't track bugs that i fix on the spot. perhaps i could, but i think that might be too heavyweight for the project's maturity level. csvbase is really only a step or two on from MVP :)

calpaterson, 11 days ago

@lewiscowles1986 @jamescooke one example of a bug that has come up several times, though, is failing to seek a file back. here's my current solution for that

imo file handling is a lot like manual memory management (because that's what it is!)

calpaterson, 11 days ago

@lewiscowles1986 @jamescooke an example use. there is a lot of this because obviously if you materialise a csv file eg as a list of dicts you timeout

calpaterson, 11 days ago

@lewiscowles1986 @jamescooke usually the whole file is being read but if you turn 50mb of csv into lists+dicts or whatever you will usually exceed memory and/or timeout

and i want to raise the limit on uploaded csv file sizes to at least 100mb

calpaterson, 11 days ago

@lewiscowles1986 @jamescooke so streaming the file is essential

as is rewinding as necessary - because various scanners run on the first part of the file, eg to infer delimiter, charset, column types, etc

calpaterson, 2 months ago

@adamchainz I'd have liked to have read this and understood the issue but the author's desire to be creatively insulting completely got in the way. Never liked this kind of tech writing and don't understand what the incentives are that make people publish stuff like this

calpaterson, 7 months ago

@adamchainz @soxofaan nice. many thanks for this, it makes a big difference and i've used it on many projects