fsfind.m

Description

fsfind is a powerful file searching utility for MATLAB. It has some notable improvements over the built-in dir command:

• results are returned as string objects
• the search depth is customizable
• the search pattern supports regular expressions
• a regex pattern can be provided for each depth of the search, which makes it possible to efficiently search very deep directory structures
• C++ MEX back-end allows it to be even faster than dir (in some cases)

Getting started

Just clone the repository and run fsfind in the command window! If you are on a UNIX machine, it will compile the MEX code prior to running the search. If you are on Windows, it will default to the non-MEX path (for some reason, the MEX version is slower on Windows). If you ever wish to override this and use the MEX version, you can call compile_mex_listfiles() to build the supporting MEX function. The correct code path will be selected inside fsfind.

Example usage

Find all files in the current directory:

files = fsfind()

Find all .m files (recursively) under the current directory:

files = fsfind(pwd, '\.m$', 'Depth', inf)

Assume we have a directory structure of the following form:

• root
  • dataset-1
    • collect_1
    • collect_2
    • ...
    • results
      • data.csv
  • dataset-2
    • collect_1
    • collect_2
    • ...
    • results
      • data.csv
  • dataset-3
    • ...
  • other-junk

We want to find all data.csv files under the file system. The brute-force way would be:

files = fsfind(root, 'data\.csv', 'Depth', inf)

However, if we know that this directory structure will be consistent, we can optimize the search by supplying a filter at each depth level like so:

files = fsfind(root, 'data\.csv', 'DepthwisePattern', {'dataset-\d+', 'results'})

This way, the search does not go inside each collect_ folder because the DepthwisePattern only includes matches against the results folder at the second depth level. Note that each pattern only needs to partially match--so in this example, dataset-\d+ could just as well be dataset and we would get the same result. Also, since we did not specify Depth but we did specify the DepthwisePattern, the Depth defaulted to one more than the length of the filter-- in this case, 3.

This concept is powerful for large filesystems. You can design fast searches on directories 10+ levels deep that would otherwise take ages using something like dir(**/*.m).