lambda expressions

[Fernando Pérez]

Gerson Kurz wrote:

> And, of course, my examples were not taken "from real life". There is
> however one nice usage of lambda: passing context.

[snip]

Quick comment. I actually need this frequently precisely for numerical 
methods, but prefer to use the alternate approach of writing the context-part 
as extra arguments. The reason is that the lambda approach generates an extra 
nested function call for each of your original function calls, and in a 
numerical algorithm this can be very expensive (particularly considering how 
expensive function calls are in Python).

A quick example:

#-----------------------------------------------------------------------------
def constrained_svdfit(x,y,sig,fix_i,fix_v,funcs,*args,**kw):
    
    # this will be a (num_params X num_datapts) matrix
    phi_base=funcs(x,*args,**kw)

    [ snipped rest of code...]


Here, funcs is a function that returns a set of fit functions evaluated at 
the array of input data x. All extra context is passed to funcs via *args and 
**kw. If on top of that I also want constrained_svdfit() to take keyword 
args, I just set up a naming convention so that they all begin with _svd_ or 
something similar, and strip them from the **kw dict before passing it to 
funcs().


Cheers,

f


PS. If by any chance the actual code is of interest to you, just email me. I 
removed the numerical details from this post but the above is taken from 
actual working code.