fmt -f <unlimited> should be supported
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
coreutils (Ubuntu) |
New
|
Undecided
|
Unassigned |
Bug Description
Strictly speaking this is an enhancement request.
fmt imposes an artificial limit on the maximum output line length controlled by the -f option, which prevents using this tool to "join" together all lines in each paragraph (for any paragraph size). This operation is necessary to prepare plain-text for import into a word processor such as LibreOffice, where "manual line breaks" (i.e. newlines in the middle of paragraphs) are undesirable. If fmt could be run with an effectively unlimited max line length, then it could be used for this purpose.
Ideally there would be a way to specify an explicitly unlimited output line length (say, -f -1).
SUMMARY:
'fmt -f 9999 file.txt' gives error "invalid width". It should accept an arbitrarily-large value (up to max 32-bit integer).
Ideally, an option would allow specifying an explicitly unlimited (or maximum) output line length.
ProblemType: Bug
DistroRelease: Ubuntu 10.10
Package: coreutils 8.5-1ubuntu3
ProcVersionSign
Uname: Linux 2.6.35-30-generic x86_64
Architecture: amd64
Date: Wed Oct 5 14:45:27 2011
InstallationMedia: Ubuntu 10.10 "Maverick Meerkat" - Release Candidate amd64 (20100928)
ProcEnviron:
PATH=(custom, user)
LANG=en_US.utf8
SHELL=/bin/bash
SourcePackage: coreutils
jimav wrote:
> Strictly speaking this is an enhancement request.
>
> fmt imposes an artificial limit on the maximum output line length
> controlled by the -f option, which prevents using this tool to "join"
You meant -w, not -f, throughout.
Thanks for the suggestion. Note that the code has this:
/* Size of paragraph buffer, in words and characters. Longer paragraphs
are handled neatly (cf. flush_paragraph()), so long as these values
are considerably greater than required by the width. These values
cannot be extended indefinitely: doing so would run into size limits
and/or cause more overflows in cost calculations. FIXME: Remove these
arbitrary limits. */
#define MAXWORDS 1000
#define MAXCHARS 5000
where MAXCHARS/2 specifies the largest width.
I.e., fmt -w 2500 works, but not 2501.
We agree that there should not be such a limit.
But the internals of fmt are not pretty -- significantly less
so than most other parts of the coreutils, and as the comment says
we cannot easily increase them arbitrarily.
In the mean time what can you do if you want truly unlimited-length
paragraphs? It's not trivial since you want to retain paragraph delimiters.
This perl command should do the trick.
It processes your input a paragraph at a time, replacing each newline
(and spaces before/after) with a single space:
perl -00ple 's/\s*\n\s*/ /g'
E.g., given this input,
1
2
3
4
1
2
3
4
5
It prints this:
$ (seq 4; echo; seq 5) | perl -00ple 's/\s*\n\s*/ /g'
1 2 3 4
1 2 3 4 5
It doesn't preserve indentation, but if you're just going to
paste it into libreoffice, that should be fine.
I've Cc'd the upstream bug-tracker, so we'll have a bug number there, too.