Stripping tags with sed?

Question

Hello,
I saw a tip at macosxhints describing how to extract text from the new MS word docx files. I thought it would be fun to come up with a shell solution, since it really just involves stripping tags from an xml file. I wanted to use sed, but being BRE challenged, I resorted to perl which seems a little heavy handed:
unzip -p some.docx word/document.xml | perl -pe 's/<[^>]>|[^[:print:]]//g'
Is it possible to do this with sed? Well it must be, I just can't get my head around it.
Cole
15 PB   Mac OS X (10.3.4)  

Answer 1

Hi Cole,
   When in doubt, ask here: sed and awk a specialty. Sed is probably a better solution in terms of speed, since there's far less overhead. Fortunately, everything in your regular expression is valid in extended regular expressions. Sed doesn't use those by default but can made to do so with the "-E" option. I could swear that it didn't used to be so and it may not be a feature universal to sed but it works perfectly with Apple's current version:
unzip -p some.docx word/document.xml | sed -E 's/<[^>]>|[^[:print:]]//g'
This command will remove everything that looks like a tag or is non-printable. Just remember that regular expressions are "greedy"; they match as much as is possible. Therefore, in a document containing the literal "less than" character, '<', would cause everything between that character and the next tag to be removed. Of course the "less than" character would doubtless be a problem in an XML document and I don't know how it's handled. Probably the easiest way would be the way I did it above, using the HTML entity, &lt;. Anyway, this isn't likely to be a common problem but you could avoid it with the following regular expression:
<[^<>]>|[^[:print:]]
It also excludes the "less than" character in the first character class. I'll bet that the "less than" character is illegal in tag names.
Gary
~~~~
   This "brain-damaged" epithet is getting sorely overworked. When we can speak of someone or something being flawed, impaired, marred, spoiled; batty, bedlamite, bonkers, buggy, cracked, crazed, cuckoo, daft, demented, deranged, loco, lunatic, mad, maniac, mindless, non compos mentis, nuts, Reaganite, screwy, teched, unbalanced, unsound, witless, wrong; senseless, spastic, spasmodic, convulsive; doped, spaced-out, stoned, zonked; {beef, beetle,block,dung,thick}headed, dense, doltish, dull, duncical, numskulled, pinhead; asinine, fatuous, foolish, silly, simple; brute, lumbering, oafish; half-assed, incompetent; backward, ********, imbecilic, moronic; when we have a whole precisely nuanced vocabulary of intellectual abuse to draw upon, individually and in combination, isn't it a little <fill in the blank> to be limited to a single, now quite trite, adjective?