I am looking for your recommendations on data structures to
consider for use in a program for efficiency, etc. . . .  I have tens
of thousands of small text files that are the output of some analysis
algorithms.  (Their number may grow significantly in the future.)

   Each file name indicates a variety of attributes of the image
files from which they were derived.  I need to read in each file and
count a few items for each but in a number of ways, according to
different attributes individually and in combination.  Then I need to
write out these counts in many different ways to a CSV output file.

   Finally, my question: What data structure(s) would you recommend
that I consider to hold all these different counts by attributes (and
combinations of attributes), so I can efficiently report it out a
bunch of different ways?

   For example, if you think of these statistics in a tree structure,
I want to be able to take each branch of the tree and quickly
determine the sum of the counts.  I also want to be able to go down to
particular leaves of the tree and sum their counts.  etc.

   Thanks in advance for any recommendations.