Repository navigation
staticcheck: misuse of slice make() and append() #112
Description
A problem that seems to occur regularly is creating a slice with a length component, then only appending to it.
Example: https://play.golang.org/p/CehTV0VFKp
Generally, the intention would be making a zero-length slice with an initial capacity, then appending to it. The effect of this occurs only at runtime and the cause can be hard to spot when debugging and/or reading the related code -- it's a very subtle detail.
There may be cases where this was done intentionally to allocate some space for other functions to set the data, or to suit a particular protocol or data structure.
However, I expect that it's exceptionally rare compared to the common misuse.
A check that observes a slice created with make(slice-type, len) followed by only append()'s to that slice could be sufficient to detect the common mistake.
Activity
mvdan commentedon May 31, 2017
What if you intentionally wanted a number of zero values at the beginning? I've needed to leave a leading
0before, for example when populating a slice that will later be used for a binary search.cgilmour commentedon May 31, 2017
That consideration was noted in the third paragraph.
mvdan commentedon May 31, 2017
I read that, but my point (which I didn't communicate very well) is that I think this would cause too many false positives.
The example was to show a not-so-rare use :)
I'm also surprised that this issue would happen regularly. I can see it happening with people who are new to Go and aren't very familiar with how
makeworks. But even then, it's fairly easy to spot that a slice has a chunk of zero elements at the beginning.[-]Misuse of slice make() and append()[/-][+]staticcheck: misuse of slice make() and append()[/+]ainar-g commentedon Sep 28, 2018
I am not new to Go, and I've committed this mistake several times over the last year. When you're tired you're just thinking "Hm, I should allocate space before appending." and you have a brain-fart and write
instead of
before appending.
I think this should be a check.
ainar-g commentedon Oct 29, 2018
FWIW,
doctor, the future go vet replacement, reports this:dominikh commentedon Oct 30, 2018
vetis starting to have more false positives than staticcheck ;)jcorbin commentedon Oct 30, 2019
Fwiw, I too am not new to go, and I still write
make([]T, 42)way too often when I meanmake([]T, 0, 42). In the case where i need to have some slots, it's usually more likemake([]T, 1, N); I've never usedmake([]T, N), followed by only appends, and meant it.