[PD-dev] "library" class names loading depth "1000" reached

[Jamie Bullock]

I noticed that a few externals I’ve been working with won’t load in Pd unless they are inside a folder with a specific name.

For example:

bsaylor/partconv~
iemlib/soundfile_info

If soundfile_info is in a folder called “iemlib’ which is in Pd’s search path then typing “iemlib/soundfile_info” in an object box loads the external

If soundfile_info is in Pd’s path but not in the “iemlib” folder, then Pd attempts to load the external but fails with:

soundfile_info: already loaded
...
soundfile_info: already loaded
maximum object loading depth 1000 reached
 soundfile_info
... couldn't create

Looking at the source code for these externals (and I guess this is a wider design pattern?), I notice that class constructor is bound to a symbol of the form “<folder>/<object>”, e.g. class_new(gensym("iemlib/soundfile_info”) …

Hence, if I type “soundfile_info” in an object box, even though Pd can find the external, it can’t actually create an instance of it because the symbol “soundfile_info” can’t be found. The object’s name is actually “iemlib/soundfile_info”.

What is the purpose of this design? It breaks the idea that “an external” is a file that can placed anywhere in Pd’s path... 

Is it something to do with libPd compatibility? Will such externals work in libPd e.g. if a patch has “iemlib/soundfile_info” and soundfile_info is compiled into the Pd binary, is this a measure to ensure that libPd finds the object?

Thanks,

Jamie


-- 
http://jamiebullock.com
@jamiebullock
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.puredata.info/pipermail/pd-dev/attachments/20150428/7355ce48/attachment.html>