Usually you should be using eachline instead of readlines, especially for
large files, because they return an iterator and does not allocate strings
for the individual lines, so that garbage collection can free the
intermediary strings.
kl. 04:19:25 UTC+1 tirsdag 11. mars 2014 skrev Peter Simon følgende:
>
> Thank-you to Tony for straightening me out (off-line). As clearly stated
> in the help, readchomp() is *supposed* to return a single string, and
> only strips newlines from the very end. What I wanted was
>
> lines = open("test.txt","r") do fid
>
> map(chomp,readlines(fid))
>
> end
>
>
> which works fine.
>
>
> Sorry for all this noise.
>
>
> --Peter
>
>
> On Monday, March 10, 2014 7:14:31 PM UTC-7, Tony Kelman wrote:
>>
>> +1 for more "doing-the-right-thing"-ness wrt annoying carriage returns.
>>
>> Semi-related: in test/spawn.jl, the tests on lines 17 and 23 fail when
>> run in a Windows command prompt but succeed from an MSYS/Cygwin terminal
>> because they pipe to whichever sort.exe is found in PATH, and lines are
>> terminated by "\r\n" in the output from Windows' sort.exe.
>>
>>
>> On Monday, March 10, 2014 5:20:27 PM UTC-7, Peter Simon wrote:
>>>
>>> Just tried to grab the lines in a text file on Windows using Julia
>>> Version 0.3.0-prerelease+1570 with the following code:
>>>
>>> lines = open("test.cut","r") do fid
>>> readchomp(fid)
>>> end
>>>
>>> lines came back to be a single string with embeded "\r\n" characters. I
>>> had expected to get an array of strings without any trailing carriage
>>> returns/newlines. Looking in the Julia source code I see that only '\n' is
>>> checked for.
>>>
>>> I work in an environment where we freely intermix Linux-style
>>> ('\n'-terminated) and Windows-style ("\r\n"-terminated) files. Most of our
>>> other tools work transparently with either file type, without any explicit
>>> need for conversion.
>>>
>>> Are there any plans to make Julia's utility functions such as chomp(),
>>> readchomp(), etc., be able to handle these various line ending conventions?
>>>
>>> Thanks,
>>> --Peter
>>>
>>
