Convert regex from pcre to sed to split strings

Question

I have a version regex which works in PCRE format while am having trouble getting this to work with sed using match groups.

Regex:

((^[[:alnum:]]+.*)-(\d+\.\d+\.\d+-VERS|\d+\.\d+\.\d+))

Input:

aaa1-bbb2-ccc3-dddd4-ffff5-1.0.0-VERS
aaa1-bbb2-ccc3-dddd4-ffff5-11.22.33-VERS
zzz1-bbb2-ccc3-1.0.1
zzz1-1.0.1-VERS

expected output: split strings and separate the version string

group2="aaa1-bbb2-ccc3-dddd4-ffff5"
group3="1.0.0-VERS"
group2="aaa1-bbb2-ccc3-dddd4-ffff5"
group3="11.22.33-VERS"
group2="zzz1-bbb2-ccc3"
group3="1.0.1"
group2="zzz1"
group3="1.0.1-VERS"

The above output work as expected here

However, trying to use the same version with sed does not work. What am I missing?

echo "aaa1-bbb2-ccc3-dddd4-ffff5-11.22.33-VERS" | sed -E 's#((^[[:alnum:]]+.*)-(\d+\.\d+\.\d+-VERS|\d+\.\d+\.\d+))#\3 \2#p'

regex syntax and features varies a lot between sed and pcre (see also https://unix.stackexchange.com/questions/119905/why-does-my-regular-expression-work-in-x-but-not-in-y) ... for ex: `\d` doesn't work in sed as you expect, you need `[0-9]`... if that alone solves your issue, please mark it as duplicate of https://stackoverflow.com/questions/14671293/why-doesnt-d-work-in-regular-expressions-in-sed — Sundeep, Jul 30 '18 at 01:30
Possible duplicate of [Why doesn't \`\d\` work in regular expressions in sed?](https://stackoverflow.com/questions/14671293/why-doesnt-d-work-in-regular-expressions-in-sed) — askb, Jul 30 '18 at 01:53
@revo No, `+` is an ERE metachar which can be enabled in multiple sed versions with `\+` or the `-E` option. — Ed Morton, Jul 30 '18 at 04:17
@askb why is there no output for the input line `zzz1-bbb2-ccc3-1.0.1`? Please add it or explain why it's missing. — Ed Morton, Jul 30 '18 at 04:18
@EdMorton Actually I was thinking about `\+` while typing that comment. `+` is part of POSIX standard. — revo, Jul 30 '18 at 05:14
Right, `+` part of the POSIX standard for EREs. See "BRE Special Characters" vs "ERE Special Characters" in the POSIX spec, http://pubs.opengroup.org/onlinepubs/9699919799/, and some sed versions (e.g. GNU and BSD/OSX) let you use EREs instead of BREs by adding the `-E` flag and GNU sed also lets you use it by preceding the ERE metachars with backslash. — Ed Morton, Jul 30 '18 at 12:08

score 0 · Answer 1 · answered Jul 30 '18 at 01:48

I think \d isn't recognised by sed. This works for me on OSX.

sed -E 's/([[:alnum:]]+.*)-([0-9]+\.[0-9]+\.[0-9]+|[0-9]+\.[0-9]+\.[0-9]+-VERS)/\1 \2/'

Input:

aaa1-bbb2-ccc3-dddd4-ffff5-11.22.33-VERS
aaa1-bbb2-ccc3-dddd4-ffff5-1.0.0-VERS
zzz1-bbb2-ccc3-1.0.1
zzz1-1.0.1-VERS

Output:

aaa1-bbb2-ccc3-dddd4-ffff5 11.22.33-VERS
aaa1-bbb2-ccc3-dddd4-ffff5 1.0.0-VERS
zzz1-bbb2-ccc3 1.0.1
zzz1 1.0.1-VERS

score 0 · Answer 2 · answered Jul 30 '18 at 01:52

0

As @Sundeep pointed out \d+ does not work with sed and should be using [0-9]+ instead.

echo "aaa1-bbb2-ccc3-dddd4-ffff5-11.22.33-VERS" | sed -E 's#((^[[:alnum:]]+.*)-([0-9]+\.[0-9]+\.[0-9]+-VERS|[0-9]+\.[0-9]+\.[0-9]+))#\3 \2#g'

answered Jul 30 '18 at 01:52

askb

6,501
30
43

1

`[0-9]+\.[0-9]+\.[0-9]+-VERS|[0-9]+\.[0-9]+\.[0-9]+` can be written as `[0-9]+\.[0-9]+\.[0-9]+(-VERS)?` which in turn can be written more concisely as `([0-9]+\.){2}[0-9]+(-VERS)?` but in reality all you need given your data is `[0-9.]+(-VERS)?`. See https://stackoverflow.com/a/51586900/1745001. – Ed Morton Jul 30 '18 at 04:30
1

ah, thks ... this saved me some time in making it compact. :) – askb Jul 30 '18 at 06:16

Ed Morton · Accepted Answer · 2018-07-30T04:28:25.490

0

Why such a complicated regexp?

$ sed -E 's/(.*)-([0-9.]+(-VERS)?)$/\2\t\1/' file
1.0.0-VERS      aaa1-bbb2-ccc3-dddd4-ffff5
11.22.33-VERS   aaa1-bbb2-ccc3-dddd4-ffff5
1.0.1   zzz1-bbb2-ccc3
1.0.1-VERS      zzz1

or:

$ sed -E 's/(.*)-([^-]+-[^-]+)$/\2\t\1/' file
1.0.0-VERS      aaa1-bbb2-ccc3-dddd4-ffff5
11.22.33-VERS   aaa1-bbb2-ccc3-dddd4-ffff5
ccc3-1.0.1      zzz1-bbb2
1.0.1-VERS      zzz1

depending on what the output should be for input zzz1-bbb2-ccc3-1.0.1.

edited Jul 30 '18 at 04:28

answered Jul 30 '18 at 04:16

Ed Morton

188,023
17
78
185

can you explain what does the `?` does? – askb Jul 30 '18 at 06:18
1

It says something is optional; so "zero or one of the preceding expression". – tripleee Jul 30 '18 at 09:31

score 0 · Answer 4 · answered Jul 30 '18 at 08:19

0

This might work for you (GNU sed):

sed -r 'h;s/^(([[:alnum:]]+-?)+)-(([[:digit:]]+\.?){3}(-VERS)?)/group1="\1"/p;g;s//group3="\3"/p;d' file

However a simpler regexp would be:

sed -r 'h;s/^(.*)-([0-9].*)/group1="\1"/p;g;s//group2="\2"/p;d' file

answered Jul 30 '18 at 08:19

potong

55,640
6
51
83

Convert regex from pcre to sed to split strings

4 Answers4