Can You Improve This Erlang Code ? (Was "Average Length Of The Sequences In A Fasta File")

Question

[Pierre Lindenbaum]

In a previous question "Code golf: mean length of fasta sequences", Eric asked for some solutions to get the average length of the sequences in a fasta file.

I tried to anwser this question using the following Erlang code:

-module(golf).
-export([test/0]).

line([],{Sequences,Total}) ->  {Sequences,Total};
line(">" ++ Rest,{Sequences,Total}) -> {Sequences+1,Total};
line(L,{Sequences,Total}) -> {Sequences,Total+string:len(string:strip(L))}.

scanLines(S,Sequences,Total)->
        case io:get_line(S,'') of
            eof -> {Sequences,Total};
            {error,_} ->{Sequences,Total};
            Line -> {S2,T2}=line(Line,{Sequences,Total}), scanLines(S,S2,T2)
        end  .

test()->
    {Sequences,Total}=scanLines(standard_io,0,0),
    io:format("~p\n",[Total/(1.0*Sequences)]),
    halt().

Compilation/Execution:

erlc golf.erl
erl -noshell -s golf test < sequence.fasta
563.16

this code seems to work fine for a small fasta file but it takes hours to parse uniprot_sprot.fasta (in fact , I pressed Ctr-C). Why ? I'm an Erlang newbie, can you improve this code ?

[Fred Fleche]

Pierre, in the mean time you should may be post your question in stackoverflow as well. Just a suggestion.

[Pierre Lindenbaum]

Fred, I'll do if I don't get an answer here :-)

[Pierre Lindenbaum]

posted on SO: http://stackoverflow.com/questions/3296855

Answer 1

[Pierre Lindenbaum]

The answer is here

Answer 2

[Istvan Albert]

A few years ago I read a blog post series on Erlang's text processing performance. Back then it seemed that the language did not have an efficient string representation.