Need help to figure out a regex to modify some files

Question

I have some files that I need to clean up some names.

for example:

GCA_940670685.1_Clostridium_sp_chr  3757330
GCA_940677205.1_Clostridium_colinum_chr 2035557
GCA_942548115.1_Aeromicrobium_sp_chr    3463989
GCA_943169635.1_Fenollaria_sp_chr   3260126
GCA_943169825.1_Varibaculum_sp_chr  4423380
GCA_943736995.1_Sporosarcina_sp_chr 3771420

And I need some like this:

GCA_940670685.1 3757330
GCA_940677205.1 2035557
GCA_942548115.1 3463989
GCA_943169635.1 3260126
GCA_943169825.1 4423380
GCA_943736995.1 3771420

I tried to use:

sed 's/_[A-Za-z]+_//gI' Terrabacteria_chr_lengths.tsv

sed 's/\w+_\w+_chr//gI' Terrabacteria_chr_lengths.tsv

find Results/Lengths/Bacteria -type f -exec sed -i 's/_\w+_\w+_chr//g' {} \;

But seems nothing is working, I think due my poor skills in the regex, ex. \w+\w+_chr.

Any suggestion would be appreciate. Thank you.

Paulo

Answer 1

The biggest issue is that + does not act as a quantifier in a sed basic regular expression (BRE) - you need to switch to extended regular expression (ERE) mode using -E or -r to use that (or change + to \{1,\} for a POSIX BRE version¹).

Beyond that, you seem to want to match a sequence of alphabetic characters and underscores after the initial underscore (but not ending with an underscore). So either:

sed -E 's/_[A-Za-z_]+//'

or

sed 's/_[A-Za-z_]\{1,\}//'

You don't need the g modifier, since you're making a single substitution per line.

---

¹ GNU sed actually supports \+ as a quantifier in BRE, but IMHO that just adds to the confusion.