Line count/patterns

Questions
1

Given: contents of a text file, which may have Unix or Windows line endings, as a byte list
Task: compute the line count
Example:

bytes={65, 13, 10, 66, 67, 13, 10}

I thought Count[bytes, PatternSequence[13,10]] might do the trick, but it gives zero. Count[bytes,10|13]/2 works, if I can know that I’m getting Windows line endings.

I don’t want an iterative approach.

2

What would you expect for the following? Or are you reasonably certain this won’t come up?

bytes={65, 13, 10, 66, 67, 10}
3

I think this works

Length[Split[bytes, #1 != 10 && (#1 != 13 || #2 == 10) &]]

It’s a little counterintuitive in this context that Split splits when the test function returns False; thus the test is

!(#1==10 || (#1 == 13&&#2!=10))

to avoid splitting on a 13 if it’s followed by a 10, because I assume you also want Mac line endings to work.

4

If you assume it’s consistent within the list (either {10} or {13,10} or {13} for every line ending), then something like

In[103]:= Max[Count[bytes, 10], Count[bytes, 13]]
Out[103]= 2

is probably pretty efficient. If you want to support a mix and match, something like this should work:

In[104]:= StringCount[FromCharacterCode[bytes], RegularExpression["\r\n?|\n"]]
Out[104]= 2

Count[bytes, PatternSequence[13,10]] doesn’t work because Count doesn’t support sequence matching, only element matching. To do that kind of thing for Mathematica expression, one typical approach is

In[109]:= Count[Partition[bytes, 2, 1, 1], {13, 10}]
Out[109]= 2

but it’s not very efficient (and wouldn’t cover the other cases anyway).

5

For those that assume the last line is terminated by a new line, here’s another approach

newlines = 1 - Unitize[bytes - 10];
formfeeds = 1 - Unitize[bytes - 13];
both = Rest[newlines]*Most[formfeeds];
Total[newlines] + Total[formfeeds] - Total[both]

I haven’t check relative efficiency.