ssort valuable? ( was Re: request for sponsoring lina32)

[Albert van der Horst <albert@spenarnc.xs4all.nl>]

Paul Wise schreef op 2017-07-04 04:29:
> On Mon, 2017-07-03 at 17:48 +0200, Albert van der Horst wrote:
>
> > To really help me you should have answered:
> >   .deb files? the purpose of packaging is to build a dsc file.
>
> Thanks for the feedback. I had assumed the packaging guide referenced
> by that page would convey that need.
> <SNIP>
The above and snipped answers really helped me along. Still
it took me some time to get through all the docs.

> I'm not sure if this is the case for ciforth+ssort, but when the choice
> is to reimplement a dependency or depend on it, I would always choose
> to depend on it. This way there is less duplication of code, which
> means less code maintenance and more shared maintenance.

The first 5 reasons amount to ssort not being of sufficient interest
and quality for Debian.

1. ssort for MSDOS has been on the internet since 1993.
  It never drew attention or was downloaded as far as I know.
2. ssort for linux was a quick and dirty port just for lina
   (It does pass the comprehensive test.)
3  It is not optimised for modern systems. The heapsort used has bad
  locality of access compared to quicksort derivates.
  A quick test reveals a factor 40 slower than gnu sort
4. Sorting using a file based merge is no longer indicated.
5. The documentation doesn't fit. Contrary to my habits, the latest
 docs contains features I wanted to add, but never did.

Then there is my personal distaste.
6. It stinks to require a 10 source, complicated build, C++
 program to support such a simple one source program.

The bottom line is that the relation with ssort might hold back a
prospective sponsor for lina.

I like the situation now better. A non-trivial task is performed by a
rather simple and straightforward lina script.
This may convince a prospective sponsor that lina is good for something.

> PS: I seem to remember thinking that ssort offers features not
> available elsewhere in Debian, so you might want to consider packaging
> it for Debian too.

I have no overview of Debian packages like you guys. Also I don't know
whether an awk or perl script couldn't accomplish the same.

The idea of ssort is that records can have multiple fields, but these
are specified as regular expression, fields may be multiple lines.
The type of comparison and priority of the fields is specified on the
command line. If that sounds interesting, one can have a look at
 http://home.hccnet.nl/a.w.m.van.der.horst/ssort.html containing 
references to
source and documentation. Contact me if you think it should be in
Debian.

Groetjes Albert

--
Suffering is the prerogative of the strong, the weak -- perish.
Albert van der Horst