Re: [COMMITTERS] pgsql: Clean up the #include mess a little.

Excerpts from Ants Aasma's message of mar sep 06 12:40:04 -0300 2011:
> On Mon, Sep 5, 2011 at 4:55 PM, Greg Stark <stark@mit.edu> wrote:
> > What I wouldn't mind seeing is a graph of all includes and what they
> > include. This might help figure out what layering violations there are
> > like the one that caused this mess. I think I've seen tools to do this
> > already somewhere.
> 
> I whipped together a quick Python script to do this. Attached is the
> Python script (requires pydot) and the result of running it on includes/.
> I didn't attach the png version of the output because it was 7MB.
> 
> If rendering all includes at once doesn't give a good overview it can
> also select a subset through traversing dependencies. For example:
> render_includes.py -i include/ \
>   --select="storage/spin.h+*,access/xlog.h+*" output.png
> 
> This will render everything that directly or indirectly depends  on those
> two headers. See --help for details.

Wow, interesting, thanks.

What this says to me is that we should do something about execnodes.h
and some other nodes file (parsenodes.h I think).

I wonder what happens if files in the same subdir are grouped in a
subgraph.  Is that possible?

-- 
Álvaro Herrera <alvherre@commandprompt.com>
The PostgreSQL Company - Command Prompt, Inc.
PostgreSQL Replication, Consulting, Custom Development, 24x7 support