it seems that the following three objects hash to the same thing (in nextflow):
- a map/hashtable with one value
- a list with that value
- value itself.
I tested this hypothesis with the following simple script:
process hello {
tag "$metaid"
input:
val(metaid)
output:
path "hello.txt", emit: hello
script:
print(metaid)
"""
echo $metaid > hello.txt
"""
}
workflow{
meta=[id:"test"]
// ch = Channel.from([meta]).map{ meta->[meta.id]}
// ch = Channel.from([meta])
ch = Channel.from([meta]).map{ meta->meta.id}
ch.view()
hello(ch)
}
and running it with -resume and changing which of the three lines in the end are not commented out. The results obtain a cache hit.
Am I missing something?
This could lead to false-positive cache hits, and hard to find bugs… I think it would be good to hash the keys and not only the values (in a map) and to include type information (to differentiate between a list containing one element and that element (behaviour could be different in the code…)