Re: [Statcvs-users] script for many projects , many repositories, by developer

[Marc l. <mar...@ya...>]

> Hi Marco,
> 
> Sounds interesting, although it is like an aggregation
> mechanism once the reports are generated right?
> 
> StatCVS could export things in XML by the way, so you may
> be able to write some code to read n XML files and do your
> aggregations that way.
> 
> 
> In any case, do send us the info / scripts. 
> 
> Thanks
> 
>


Hi Benoit,

You are right, it aggregates projects.

I think it is a kind of zoom out, for managers. They are interested more on what their developers are doing recently.
Developers, on the other hand ,zoom in. Down to differences between file versions. Different audience, at least here where I work.

I think xml are better then parsing html files,  but I have to do it in a hurry and didnt know how to parse xml from bash (learning groovy sh right now).

BTW,it would be cool if we put the cvs info in a little DB (H2 DB comes to mind) so that we could select our report away. See below for some crazy ideas.

Im sending the hacks (err... scripts)  to save them on internet, Im always formating my drives here :-). 

 Thats it. Hope they are useful.
Thanks

Marco "Linux" Antonio de Sousa
linux/java developer, Brazil 


 
SCRIPT DETAILS

One script try to  find any user_*.html file and generate a CSV file based on that:
Eg:
projectM/user_marc.html
dir/projectN/user_marc.html ; user_ann.html

Generates

csv.csv:
projectM marc
dir/projectN marc,ann

marc.csv:
projectM
dir/projectN

ann.csv:
dir/projectN

It is pushing the limits of my (limited) shell programming capabilities :-).
A DB here would be *way* better.
From those CSV files, another script generate the html index.

The scripts works well with my cvs server, but YMMV.
I wrote them with maintainability in mind, not efficiency. 
Thats why I used so many functions.They make  it easier to test and separate concerns.
 
It needs: bash (and find,date, cron, etc). I think any linux box will do. And java for statcvs.


INSTALL/USE
Edit the scripts and configure  java, http dir, projects/repos;
Run avery(day|week) with cron;
browse to : http://your.server/http-dir/
Enjoy.

Put them on cron (user must have cvs checkout rights):
$ crontab -l
HTTP_DIR=/opt/IBM/HTTPServer/htdocs/en_US/cvs-stats/

4 4 * * * /opt/cvs-stat/cvs-stat.sh > /tmp/cvs-stat.log 2>&1 ; mv /tmp/cvs-stat.log $HTTP_DIR ;  /opt/cvs-stat/makeIndex.sh > $HTTP_DIR/makeIndex.log 2>&1

In this example , run every day at 04:04 am (every week should do it, but the machine is idle anyway).

SOME CRAZY IDEAS
statcvs could insert data on a local DB (say, H2 db) .
That way we could ask things like:
select * from projects
select users from project where project like foo
select  commit from project order by date
select  project from repositories where projName like "myProj"
select commit-date from projects order by developer

The scripts could start a local web server (jetty) to avoid setup an apache server.


      ____________________________________________________________________________________
Veja quais são os assuntos do momento no Yahoo! +Buscados
http://br.maisbuscados.yahoo.com