[SOLVED] sed - eliminate punctuation at start and/or end of each line

danielbmartin · 01-20-2012, 02:27 PM

Eliminating punctuation marks everywhere in a file is easy enough.
I want to eliminate punctuation marks only at the start (or end) (or start and end) of each line.
How is this done?

I've tried constructs such as

Code:

^[:punct:]

and

Code:

[:punct:]$

to no avail.

Daniel B. Martin

Nominal Animal · 01-20-2012, 02:58 PM

Note that character classes only work within a definition, i.e. use [[:punct:]] or [[:punct:][:space:][:cntrl:]] for example. The classes depend on locale ( LANG, and LC_ALL or LC_CTYPE environment variables). Which characters do you consider punctuation? Or is the locale definition okay by you?

For me,

Code:

sed -e 's|^[[:punct:]]*||; s|[[:punct:]]*$||;' -i file-to-edit

works fine using GNU sed 4.2.1.

jthill · 01-20-2012, 03:04 PM

Please post how you have been using those constructs ... oh. a-z0-9 in [a-z0-9] names two ranges of code points. [:punct:] in [[:punct:]] names a set of code points. Your [:punct:] names the char class consisting of :, p, u, n, c, and t. You want the doubled brackets.

(edit: yup, n.a. said it.)

danielbmartin · 01-20-2012, 03:15 PM

[QUOTE=Nominal Animal;4580122]

Code:

sed -e 's|^[[:punct:]]*||; s|[[:punct:]]*$||;' -i file-to-edit

Perfect!

Quote:

Originally Posted by Nominal Animal

Which characters do you consider punctuation? Or is the locale definition okay by you?

Okay by me!

Thanks... and we can mark this one SOLVED!

Daniel B. Martin

danielbmartin · 01-20-2012, 04:43 PM

(null)