Direction of KDevelop4

[Kuba Ober]

On Thursday 22 September 2005 09:04, Amilcar do Carmo Lucas wrote:
> On Thursday 22 September 2005 14:36, Kuba Ober wrote:
> > On Wednesday 21 September 2005 16:07, Amilcar do Carmo Lucas wrote:
> > > The hand written part of the KDevelop website has
> > > 550 000 lines of code.
> > > 250 000 links.
> >
> > I'm dumb this morning. What are those numbers? I mean, who wrote half a
> > million lines of code for the website?
>
> I committed 60% of them. (So says the CVS log) :)

Code as in php logic, or code as in html? Or a mix of the two?

> > What does the code do?
>
> Builds a very friendly site with ZERO regressions.

How can one person write the code equalling in size to like a fifth of the 
linux kernel? Apparently in just a couple years? And, unless I'm mistaken, 
without making it a full-time job with double overtime? Or I just plain don't 
know what I'm talking about. Or the lines are very very short ;) But 
seriously, how much redundancy is in the lines that you mention in those 
counts?

I wouldn't call html marked-up content code. I'd call code the php and sql 
logic that builds the friendly site, that's for sure.

> > And who put
> > quarter of a million links there?
>
> Me helped a lot! and PHP and MySQL and python and XSLT and bash.

Is it just me, or is it ridiculous to count links in the generated html?

I mean, I don't want to belittle anything that you did, the site works great, 
but the numbers seem just pretty much meaningless without some context 
accompanying them. What you did seems to me akin to saying that google's site 
has however many quintillion links, or whatever other big number you obtain 
by combining various ways one can query their database.

> > Where are all those links?!
>
> Distributed in the 1800 handwritten webpages (in 15 languages).

Let's get sane here. How many of such pages are incremental revisions over 
historical pages that are just kept? Not that it's bad to keep them, of 
course. It's akin to counting every single revision of every possible page. 
I've looked at a small (15k loc) C++ project that I maintain and it looks 
like even though there are like 100 files in the baseline, there have been 
about 5000 file revisions in the last 4 years. Some of the files exist no 
more and even their functionality might not exist anymore (say because Qt 
took over that). That still doesn't make my project 5000 files big. It's 100 
files small, all right.

Again, I do appreciate all the history-preserving aspects of the website and 
all the logic that works to keep things rolling smoothly, but I don't think 
it's sane at all to count anything but content related to current stable 
version and to head. And then counting translations of a page is equally 
misleading, as there's really no new content added by translation. That's 
like adding up wikipedia article counts from wikipedias in various languages 
without taking into account how much overlap there is between those due to 
translation. Of course original articles in other languages do count, but not 
all of them are original.

I'd say that it's sane to say there are x pages, N% of which are translations, 
P% of which relate to versions other than current stable or HEAD. I bet N and 
P will not be insignificant.

I only question the metrics you've used. Somehow small numbers don't seem to 
work for people anymore :(

> > (I don't have one handy). Does a single kdevelop
> > version generate 120mb worth of html output from doxygen? That's scary ;)
> yes.
 
Which still means that in reality there can be no more handwritten 
documentation than say 80% of the size of the source tree, and that'd be for 
a very well documented project. Of course assuming that all doxygen input is 
kept in the source tree. So, what was the size of the head (or latest stable 
release) again?

I just think that counting the machine generated output from doxygen as 'size 
of the documentation' is completely pointless, unless you want to show off 
disk space usage on the web server. It has little to do with how much real 
work went into the actual project. My little 15k line C++ project generates 
some 10mb worth of doxygen output. I'd consider myself crazy if I used that 
10mb number for anything but lamenting over how space-inefficient it is to 
keep doxygen output around ;)

Cheers, Kuba