Hi Brian,
On 2023-08-28 20:24, Brian Inglis wrote:
> On 2023-08-28 06:17, Alejandro Colomar wrote:
>> Hi Brian,
>>
>> On 2023-08-22 01:45, Brian Inglis wrote:
>>> I am in favour of all punctuation being treated as word spaces and sorting
>>> "cat ..." before "cat..." but find the real orders more evocative and easier to
>>> decide about than examples.
>>
>> Here's an excerpt of how treating - and _ as spaces looks like. I think
>> it's a reasonable order. Should I apply that diff?
>>
>> Cheers,
>> Alex
>>
>> $ git diff
>> diff --git a/scripts/sortman b/scripts/sortman
>> index a8f70bab5..6d1d92f09 100755
>> --- a/scripts/sortman
>> +++ b/scripts/sortman
>> @@ -9,7 +9,7 @@ sed -E '/\/intro./ s/.*\.([[:digit:]])/\10\t&/' \
>> | sed -E ' s/\t(.*)/&\n\1/' \
>> | sed -E '/\t/ s/\.[[:digit:]]([[:alpha:]][[:alnum:]]*)?\>.*//' \
>> | sed -E '/\t/ s/\/[_-]*/\//g' \
>> -| sed -E '/\t/ s/[_-]/_/g' \
>> +| sed -E '/\t/ s/[_-]/ /g' \
>> | sed -E '/\t/ {N;s/\n/\t/;}' \
>> | sort -fV -k1,2 \
>> | cut -f3;
>> $ touch man8/ld-z.8
>> $ touch man8/ld.8
>> $ find man8 | ./scripts/sortman
>> man8/intro.8
>> man8/iconvconfig.8
>> man8/ld.8
>> man8/ld-linux.8
>> man8/ld-linux.so.8
>> man8/ld-z.8
>> man8/ld.so.8
>> man8/ldconfig.8
>> man8/nscd.8
>> man8/sln.8
>> man8/tzselect.8
>> man8/zdump.8
>> man8/zic.8
>> man8
>
> Looks better,
Thanks, I've applied and pushed the patch.
> but should your sort *key* field instance also drop the section
> suffix (already in prefix)
It is already dropped. Am I understanding it correctly?
Here's a debug patch to view the sort key field:
diff --git a/scripts/sortman b/scripts/sortman
index 6d1d92f09..e690f23ea 100755
--- a/scripts/sortman
+++ b/scripts/sortman
@@ -12,4 +12,5 @@ sed -E '/\/intro./ s/.*\.([[:digit:]])/\10\t&/' \
| sed -E '/\t/ s/[_-]/ /g' \
| sed -E '/\t/ {N;s/\n/\t/;}' \
| sort -fV -k1,2 \
+| tee /dev/tty \
| cut -f3;
And here's how it looks with man8 (plus the dummy files):
$ find man8 -type f | ./scripts/sortman
80 man8/intro man8/intro.8
81 man8/iconvconfig man8/iconvconfig.8
81 man8/ld man8/ld.8
81 man8/ld linux man8/ld-linux.8
81 man8/ld linux.so man8/ld-linux.so.8
81 man8/ld z man8/ld-z.8
81 man8/ld.so man8/ld.so.8
81 man8/ldconfig man8/ldconfig.8
81 man8/nscd man8/nscd.8
81 man8/sln man8/sln.8
81 man8/tzselect man8/tzselect.8
81 man8/zdump man8/zdump.8
81 man8/zic man8/zic.8
man8/intro.8
man8/iconvconfig.8
man8/ld.8
man8/ld-linux.8
man8/ld-linux.so.8
man8/ld-z.8
man8/ld.so.8
man8/ldconfig.8
man8/nscd.8
man8/sln.8
man8/tzselect.8
man8/zdump.8
man8/zic.8
There are no suffixes in the second field.
> and also treat "." as space?
I'had been thinking about it, but didn't make an opinion.
Since they are rare, I think making them stand out a little bit
by having a special order rather than just being mixed with the
underscores would make sense. But I'm open to change that.
> Where would you expect to see ld.so?
Not sure.
>
> Also, in `sed`, instead of cloning the line, at the start of a series of
> executions, make them all into a single inline command script, start with `h` to
> *hold* the input line, and end with `G` instead of `N` to append '\n' then the
> held line, convert to `\t`, drop the braces, and you can skip the then redundant
> tests, something like the following should get you close (tried it earlier, now
> sadly already gone from history):
>
> | sed -E '
> h
> /\/intro./ s/.*\.([[:digit:]])/\10\t&/
> s/\.[[:digit:]]([[:alpha:]][[:alnum:]]*)?\>.*//
> s/\/[_-]*/\//g
> s/[_-]/_/g
> s/[_-]/ /g
> G
> s/\n/\t/
> ' \
> | ...
I prefer having many one-liners for a few reasons:
- Not everybody knows what h and G do. I did't. And I will
soon forget. In contrast, my implementation has nothing
rare in it.
- I can inspect the contents at each of the steps easily by
adding a line with `| tee /dev/tty \`, for debug purposes.
In general, I avoid having large scripts in other languages.
I prefer piping many one-liners, even if it might be less
efficient (but it uses more cores, so it might end up being
faster; I've seen such things happen already many times).
Cheers,
Alex
--
<http://www.alejandro-colomar.es/>
GPG key fingerprint: A9348594CE31283A826FBDD8D57633D441E25BB5
Attachment:
OpenPGP_signature
Description: OpenPGP digital signature
