4

I have a Perl script which requires the following arguments:

count.pl OUTPUT_FILE INPUT_FILE

What I want to do is to use process substitution instead of specifying INPUT_FILE explicitly, for instance:

count.pl count.txt <(cat test.txt)

However, it seems that the script does not receive any input from process substitution.

What I'm doing wrong?

Improve this question
6
  • What is the script doing with that second argument? Commented Sep 30, 2020 at 10:44
  • @Kusalananda INPUT_FILE is the text file which is the input to the script. Script process the INPUT_FILE and writes to the OUTPUT_FILE. Commented Sep 30, 2020 at 10:57
  • How does it process the input file? Does it read the file linearly as a stream, or does it try to search back and forth in the data? Commented Sep 30, 2020 at 11:00
  • @Kusalananda Script process file linearly (line 523). Commented Sep 30, 2020 at 11:06
  • Do you get an error message from the script? Commented Sep 30, 2020 at 11:10

1 Answer 1

Reset to default
6

The script uses -T the-source-file to determine whether the file contains text.

That -T operator reads the start of the file and uses heuristics to determine whether it looks like text or not. But if the file is a pipe like in the <(...) case, then what it has read cannot be read again.

You can reproduce with:

$ perl -e '$file = shift; -T $file; open F, "<", $file; print while <F>' <(seq 1867)

1861
1862
1863
1864
1865
1866
1867

See how -T has consumed 8KiB from the output of seq which are no longer there for the while <F> loop to read.

Here, you could use zsh and its =(...) form of process substitution that uses a temporary file instead of a pipe. Or (... | psub -f) in the fish shell. Or remove that call to -T from that script. Or rewrite that script so it works with <<>>/<> (so it would also work on stdin when not passed any file argument) and outputs on stdout. I don't really see the point of it taking file names as arguments as it processes all of them in a stream fashion.

Improve this answer
4
  • @user431397, for regular files, it's -f. -T works as documented, I see no bug in there. Commented Sep 30, 2020 at 17:42
  • @user431397, the script uses <HANDLE> on each file opened in turn, which like <> or <<>> reads one full record at a time. My suggestion of switching to <>/<<>> will make no difference in that regards. Commented Sep 30, 2020 at 17:46
  • @user431397, you'd have the same problem with -T "/dev/fd/x" even if x is a fd open on a regular file on systems other than Linux/Cygwin which would also have the side effect or moving the cursor within the file (-T could be modified to seek back in that case I suppose). -T would also be fine on other type of seekable files like block devices. The documentation is already clear that it does read from the file. Maybe it could give a warning that it makes little sense to use it on non-seekable files. Commented Sep 30, 2020 at 17:55
  • @user431397 Sorry you're right. I overlooked that stat() would not return +regular+ on /dev/fd/x on non-Limux (even if fstar() would on a FD opened on it) Commented Oct 1, 2020 at 5:41

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.