String arrays with linear time lookup

Question

I am doing string processing in Matlab and I usually use cell arrays to store individual words in the text

Example:

a = {'this', 'is', 'an', 'array', 'of', 'strings'}

For searching for the word 'of' in this array I loop through the array and check each individual element against my word. This method does not scale since if I get a large dataset my array a will grow large and looping through elements is not wise. I am wondering if there is any more smart way, perhaps a better native data structure in Matlab, that can help me run this faster?

Relevant read http://blogs.mathworks.com/loren/2006/12/20/finding-strings/ — oleksii, Jan 24 '12 at 20:12
Possible duplicate of http://stackoverflow.com/questions/983163/map-function-in-matlab/983239#983239 — Andrey Rubshtein, Jan 24 '12 at 22:21

gnovice · Accepted Answer · 2012-01-24T20:27:09.480

A map container is one option. I don't know what specific sort of string processing you intend to do, but here's an example for how you can store each string as a key which is associated with a vector of index positions of that word in a cell array:

a = {'this', 'is', 'an', 'array', 'of', 'strings', 'this', 'is'};

strMap = containers.Map();  %# Create container
for index = 1:numel(a)      %# Loop over words to add
    word = a{index};
    if strMap.isKey(word)
        strMap(word) = [strMap(word) index];  %# Add to an existing key
    else
        strMap(word) = index;  %# Make a new key
    end
end

You could then get the index positions of a word:

>> indices = strMap('this')

indices =

     1     7    %# Cells 1 and 7 contain 'this'

Or check if a word exists in the cell array (i.e. if it is a key):

>> strMap.isKey('and')

ans =

     0    %# 'and' is not present in the cell array

String arrays with linear time lookup

1 Answers1