What is the equivalent most efficient way to tail -n 1 file.csv in matlab

For context we write to the .csv every test collect with general info and I have some analysis tools that grab some information from that file very consistently. I was just curious if there is something better than:

system('tail -n 1 file.csv')

built in to read the last line without traversing the entire file.

Answer 1

[Walter Roberson]

There are a few cases:

If there is a maximum line length that the final line is certain not to exceed, and the characters are restricted to 8 bit, then

fid = fopen('file.csv', 'r');
fseek(fid, -(MAXIMUM_LINE_LENGTH+1), 'eof');
buffer = fread(fid, '*char*1');
%get rid of carriage returns. Get rid of trailing newlines
buffer = regexprep(buffer, {'\r', '\n+$'}, {'',''});
pos = find(buffer == newline, 1, 'last');
lastline = buffer(pos+1:end);

If there is a maximum line length that the final line is certain not to exceed, and the characters might be multibyte, then

fid = fopen('file.csv', 'r');
fseek(fid, -4*(MAXIMUM_LINE_LENGTH+1), 'eof');
buffer = fread(fid, '*char');
%get rid of carriage returns. Get rid of trailing newlines
buffer = regexprep(buffer, {'\r', '\n+$'}, {'',''});
pos = find(buffer == newline, 1, 'last');
lastline = buffer(pos+1:end);

If there is no maximum line length that the final line is certain not to exceed, then you have two choices:

• you can read everything from the beginning of the file, and throw away all except the last line
• you can start at the end of the file and seek backwards by chunks and reading each chunk, each time looking for a newline in the chunk, until eventually you find the newline or you get to the beginning of the file

Note: I coded so that trailing empty lines on the file are ignored. Otherwise you run into an existential question: if the file ends in a newline, then does that mean that there is a trailing empty line "following" the newline, or does the trailing newline mark the end of an existing line? For both Windows and Unix, the API answer is that trailing newline implies empty line afterwards: newlines are officially line seperators rather than line terminators.

[Peter]

Thank you for gracing me with an answer Walter! I ended up implementing this portion and optimizing the further parsing for my applicaiton. By the grace of god, my data is just utf-8 and there is a pretty static character length. Good catch on the trailing endline, its news to me.

fid = fopen('file.csv', 'r');
fseek(fid, -(MAXIMUM_LINE_LENGTH+1), 'eof');
buffer = fread(fid, '*char');

Have yet to profile everything but agnostic OS analysis tools are something that I strive for in my repo.

[Walter Roberson]

For utf8 you need 4 times the maximum line length, as potentially the last line might consist entirely of Linear A or other characters that require 4 bytes to represent in utf8

[Walter Roberson]

Unless, that is, you have sequences of emoji. Apparently fully specifying some emoji takes 7 bytes.

[Peter]

That was a typo I did in fact use -4(x+1)