On Thu, 7 Aug 2025, The Wanderer wrote:
On 2025-08-07 at 22:17, Tim Woodall wrote:
iHi All,
(Typed, rather than cut/pasted so apologies if there's a typo)
This is part of a sed command in a bash script.
sed 's/["'"'"']//'
This matches " or '. It works.
But it's 'unreadable' and almost impossible to find the typo if it's
wrong.
Is there a better way of writing it? the obvious '["\']' doesn't
work.
That depends on your threshold for "better".
The problem, of course, is the overlap in characters between those
needed for the s-expression and those parsed by the shell.
My best stab at it, which is only arguably better, looks like:
sed 's/["'\'']//'
I like that. Thanks!
Alternatively you could probably do
sed s/\[\"\'\]//
I didn't mention it but this hits the same snag as "[\"']", the real
expression is much more complex, approximately:
s/.*\$ref["']\{0,1\}: *["']\([^"']*\)["'].*/\1/
which gets messy with \\\$ etc malarky without the single quotes.
Due to "official incompetence" I'm forced into scraping a malformed yaml
and json openapi specification from the web so that I can then fixup
things like illegal tabs in yaml, invalid trailing comnas in json, to
the point that it can actually be loaded by python yaml/json modules.
This is a crude 'first pass' downloader that does just enough to
download all the files I need.
Fortunately, at least so far, I've not had to cope with multiple $ref on
a single line :-)
Tim.
