Subsetting a dictionary

[Eric Hagemann]

Rather than copy the entire dictionary, why not create a new empty one and
only add the items that match you contraint.  Seems like making the copy of
all items and then throwing part of them away is extra work.  Am I missing
something ?

Cheers

"Will Newton" <will at nospam.misconception.org.uk> wrote in message
news:993dkm$ttr$1 at news6.svr.pol.co.uk...
>
> I need to get a subset of a dictionary:
>
> D = Dictionary of objects (in this case, training examples for a decision
> tree)
>
> D2 = {D | d in D, d under some constraint}
>
> I need to be able to do this, but as it is temporary I do not want to
> destructively remove examples, so:
>
> D.subset(c)
> will return a new dictionary object that is equivalent to D under
> constraint c.
>
> I have code that does this, but it's very slow, consuming typically 75% of
> program runtime - it is called many times, which seems to be unavoidable
> without modifying the algorithm (ID3) too far.
>
> I currently have:
>
> class TrainingData:
>          data = {}
>         def subset(self, attr, value):
>                 # Copy semantics
>
>                 td = TrainingData()
>
>                 td.data = self.data.copy()
>
>                 for item in td.data.keys():
>                          if item[attr] != value:
>                                    del td.data[item]
>
>                 return td
>
> Can anyone see how I can speed this up?
>
> I could conceivably do this with a list and use filter, I'll see how that
> works.