felipelouza / gsa-is Goto Github PK

I need to build a suffix array on multi-character tokens (basically 'words') rather than characters. In my first attempt I created a sorted 'alphabet' of unique tokens from the input, and assigned each an integer, then passed the string of these integer indexes, joined with a separator into the algorithm. This works, but require lots of string joins/splits, and there are additional suffixes computed that start with the separator which is not ideal.

I'm wondering if there a better way to do this, perhaps computing the SA directly on the array rather than on a string ??

A toy example given the input array [ "if", "it", "is", "true", "it", "is", "true"].

[
    0: if it is true it is true
    1: is true
    2: is true it is true
    3: it is true
    4: it is true it is true
    5: true
    6: true it is true
]

moved from: felipelouza/gsufsort#5

Implementation causes a segmentation fault when the input contains at least one empty string.

Example input

In your notation a$b$$d$#.

Expected output

Suffixes should be sorted as usual + the document index in DA, empty-suffix pointer in SA and a 0 in the LCP array.

I reckon:

i       SA      DA      LCP     suffixes
0       8       4       0       #
1       1       0       0       $b$$d$#
2       3       1       0       $$d$#
3       4       2       0       $d$#
4       6       3       0       $#
5       0       0       0       a$b$$d$#
6       2       1       0       b$$d$#
7       5       3       0       d$#

Actual output

Segmentation fault.

Workaround

Add a dummy token to all empty strings.