From: Robert Klemme <shortcutter@...>
Date: 2008-12-23T01:01:38+09:00
Subject: Re: get first and last line from txt file - how?

On 21.12.2008 19:16, Thomas Preymesser wrote:
> 2008/12/21 Robert Klemme <shortcutter@googlemail.com>:
>> On 20.12.2008 17:46, Thomas Preymesser wrote:
>>> 2008/12/20 Yaser Sulaiman <yaserbuntu@gmail.com>:
>>>> I'm just wondering..
>>>> Let's say that we only need to read the last line. Can we do that without
>>>> reading the other lines?
>>> Yes. Position your file pointer to the last byte in a file, read and
>>> collect backwards each byte until you find a newline character (or the
>>> first byte of the file). This is the last line.
>> You have to admit that this approach is rather inefficient.
> 
> Really?
> 
> I did a comparision of your code and my idea:
> 
> $ time ruby r.rb input
> 111111111111111111111111111111
> 999999999999999999999999999999
> 
> real	0m0.053s
> user	0m0.000s
> sys	0m0.004s
> 
> $ time ruby t.rb input
> 999999999999999999999999999999
> 
> real	0m0.043s
> user	0m0.004s
> sys	0m0.000s
> 
> the first result is your code, the second is mine.

Did you make sure that no OS disk buffering distorts this result?  I 
suggest to include both variants in a single script, execute each 
variant multiple times in a loop and use Benchmark#bmbm.

> I did the tests with a test file with almost 8,000,000 lines.
> 
> My q&d code:
> 
> f=File.open("input")
> pos = 2

This opens the door for character loss of the last line under certain 
conditions.

> f.seek(-pos, File::SEEK_END)
> c = f.getc
> result = ''
> while c.chr != "\n"
>   result.insert(0,c.chr)
>   pos += 1
>   f.seek(-pos, File::SEEK_END)
>   c = f.getc
> end
> f.close
> 
> puts result

Also, this code is not equivalent to mine as it does not output the 
first line - which you can nicely see from the console output shown above.

Please keep also in mind, that my code tries to do some error checking 
which avoids printing the line from a single line file twice (although 
that bit is a slightly flawed, I'll leave that debugging task as 
exercise for the reader).

A final remark: using the block form of File.open is always safer.

Cheers

	robert


-- 
remember.guy do |as, often| as.you_can - without end