bash: how do I concatenate the output of two commands so that I can pipe them to a third?

Question

$ hg status

and

$ hg status --ignored

give very similar outputs. I'd like to concatenate them so I can feed them to awk, as if there were an hg status --all (or svn's svn status --no-ignore)

I'm thinking something like:

$ echo "$(hg status)" "$(hg status --ignored)" | awk  ' ( $1 == "?" ) || ( $1 == "I") { print $2 }' | xargs rm -r

to make a 'make very clean indeed' command, but it seems to occasionally leave a file behind, perhaps because a newline goes missing or something.

Solution 1

You can use a subshell:

( hg status; hg status --ignored ) | awk '( $1 == "?" ) || ( $1 == "I") { print $2 }' | xargs rm -r

Solution 2

Use curly braces to group commands:

$ { echo first line; echo second line; } | grep "line"
first line
second line

(Posted as an answer from camh's comment)

Solution 3

You can use the rest of the hg status flags to show what you really want:

hg status -uriamn

That shows unknown files (u), removed files (r), ignored (i), added (a), modified (m) and does so without showing the status prefix.

Solution 4

This works for me:

echo $(a)$(b)

if you add "" you can add delimiters eg.:

echo "$(./gethostname.sh)|($(./getip.sh);"

I use this on Openwrt to broadcast my ip settings:

echo "$( uci get system.@system[0].hostname )|$( ip addr | grep inet | grep br-lan | cut -d ' ' -f 6 | cut -d '/' -f 1 );"  | socat - UDP-DATAGRAM:255.255.255.255:4999,broadcast ;

View more solutions

Comments

[Fred Foo]

If you replace rm with echo (or skip the last pipe), then what is the output and what should it have been? Also, there might be permission issues where rm refuses to delete a file.

[Kent]

+1 I just wrote down "why not try hg st -iu|awk ...?" and saw your answer shows up...

[camh]

There's no need to use a subshell (another process). Instead you can use braces to group: { a; b; } | c

[Ben Moss]

And I would have posted sooner, but I was trying to make an amusing anagram out of the flags ;-)

[Mark Fox]

I see no meaningful speed difference time { git branch; git branch -r; } > /dev/null; time (git branch; git branch -r) > /dev/null in fact subshell is consistently faster on my system — so @camh, what is is the advantage of a group?

[camh]

@MarkFox: It's likely measurement error. The group has one less fork() than the subshell. It is very hard to measure a single run of things like this. You want to run it 1000 times or more to get a good feel for the performance. When I do that, the group comes out at 1.69s and the subshell at 2.36s, which is about 40% slower.

[Aaron Wallentine]

Excellent, thank you. This is what I was looking for, trying to concatenate headers to pass to sendmail -t along with output of another command.