Thread: [Dev-C++] Choose a word from a .txt file

Dear users,
 
How can I write a script to look a word in a .txt file with two columns (one is the word and the other is the answer I need) and give an answer which is in the second column. In other words, a user types a word, then the script match it with a word in a .txt file (first column) and gives the answer (second column).
 
Thanks in advance!
 
Alex.
_________________________________________________________________
Cansado de espaço para só 50 fotos? Conheça o Spaces, o site de relacionamentos com até 6,000 fotos!
http://www.amigosdomessenger.com.br

Wouldn't regular expression search help you? May be better solution
could be there...

check for regex in www.boost.org


On Tue, 2008-06-03 at 01:24 -0300, Alexsandro Meireles wrote:
> Dear users,
>  
> How can I write a script to look a word in a .txt file with two
> columns (one is the word and the other is the answer I need) and give
> an answer which is in the second column. In other words, a user types
> a word, then the script match it with a word in a .txt file (first
> column) and gives the answer (second column).
>  
> Thanks in advance!
>  
> Alex.
> 



______________________________________
Scanned and protected by Email scanner

No regular expressions if he is looking for an exact match. Regular
expressions are there to look for patterns. Normal strcmp()/strncmp() is
faster when looking for exact matches.

/pwm

On Tue, 3 Jun 2008, Lloyd wrote:

> Wouldn't regular expression search help you? May be better solution
> could be there...
>
> check for regex in www.boost.org
>
>
> On Tue, 2008-06-03 at 01:24 -0300, Alexsandro Meireles wrote:
> > Dear users,
> >
> > How can I write a script to look a word in a .txt file with two
> > columns (one is the word and the other is the answer I need) and give
> > an answer which is in the second column. In other words, a user types
> > a word, then the script match it with a word in a .txt file (first
> > column) and gives the answer (second column).
> >
> > Thanks in advance!
> >
> > Alex.
> >
>
>
>
> ______________________________________
> Scanned and protected by Email scanner
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by: Microsoft
> Defy all challenges. Microsoft(R) Visual Studio 2008.
> http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
> _______________________________________________
> Dev-cpp-users mailing list
> Dev...@li...
> TO UNSUBSCRIBE: http://www23.brinkster.com/noicys/devcpp/ub.htm
> https://lists.sourceforge.net/lists/listinfo/dev-cpp-users
>

> No regular expressions if he is looking for an exact match. Regular
> expressions are there to look for patterns. Normal 
> strcmp()/strncmp() is
> faster when looking for exact matches.

No not at all. You explained an algorithms to look in the middle, then depending on less or greater in the lower half or in the upper half etc. That needs a lot of lookups if the file is huge.

A regex with the multiline option can find the answer with one line.
/^WordToFind\s*?(.*)/m
In the caption(1) you will find the answer, if there is an answer.

Regards, BM

The question here isn't the number of source lines to perform the job.
Why? Because you don't need any source lines at all if you just use normal
unix-available tools - you can get an answer directly on the command line.

But this is a C/C++ mailling list, since it is about Dev-C++. So the
question should then be: How to actually perform the job (instead
of letting existing command-line tools do it) in C/C++ source code.

If you have a nail: should you use a hammer or a sledge hammer? Or
maybe you should blast it into the pnak with dynamite? The Hilti nail-guns
goes the explosive way, but with the disadvantage that it doesn't leave
much for a DIY guy. Calling a magic reg-exp library doesn't tell someone
what is happening.

Reg-exp is cool, but regular expressions are meant to perform pattern
matching. You don't have much need for any pattern matching in this case -
and more importantly, you are ignoring how to solve the problem since you
let someone else solve it for you.

Next step: You compalin about my tree algorithm.I did suggest both a
linear approach and a tree algorithm, mentioning the tree algorithm for
large sorted files. If you have one billion words, the text file may be
100 billion bytes large. Guess how fast your reg-exp scan of the full file
would be. Guess what would happen when a tree algorithm cuts it in two for
each iterative step? A naive implementation would solve for one billion
words in max 30 tests. A slightly better approach (switching from tree to
linear when you have narrowed it down to a block of maybe a couple of kB
of data - matched to n * block size for file system or memory subsystem.

Now, assume 20MB/s for the regexp solution, and 0.1s/scan for the tree
search. On average, the regexp would then need 25 000 seconds on average
(twice in worst-case), while the tree search would take <= 3 seconds
worst-case. Hence, the simplest code doesn't always win - you must know
what your requirements are before you can select a good (or almost always
a "good enough") algorithm.

Anyway - please compare runtime and size of application binary for a
compiled regexp solution, as compared with the following:

void scan(char* my_word) {
    char tmp_match[MAX_WORD_LEN+10];
    char linebuf[MAX_LINE_LEN],*p1,*p2;
    int match_len,found = 0;
    FILE *f;

    match_len = strlen(my_word);
    strcpy(tmp_match,my_word);
    strcpy(tmp_match+match_len++," ");
    f = fopen(fname,"rt");
    if (f) {
        while (fgets(linebuf,sizeof(linebuf),f)) {
            if (!strncmp(linebuf,tmp_match,match_len)) {
                p1 = linebuf + match_len;
                while (*p1 == ' ') p1++;
                p2 = strchr(p1,'\n');
                if (p2) *p2 = '\0';
                printf("Answer for %s: %s\n",my_word,p1);
                found = 1;
                break;
            }
        }
        fclose(f);
        if (!found) printf("No match found for %s\n",my_word);
    } else {
        printf("Failed opening text file\n");
    }
}

Note: A real implementation should call setbuffer() or setvbuf() or
alternatively use ioctl() or similar and specify one-pass read-only scan
without seek. But that isn't part of a search algorithm.

/pwm

On Tue, 3 Jun 2008, "Münt, Bernd" wrote:

> > No regular expressions if he is looking for an exact match. Regular
> > expressions are there to look for patterns. Normal
> > strcmp()/strncmp() is
> > faster when looking for exact matches.
>
> No not at all. You explained an algorithms to look in the middle, then depending on less or greater in the lower half or in the upper half etc. That needs a lot of lookups if the file is huge.
>
> A regex with the multiline option can find the answer with one line.
> /^WordToFind\s*?(.*)/m
> In the caption(1) you will find the answer, if there is an answer.
>
> Regards, BM
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by: Microsoft
> Defy all challenges. Microsoft(R) Visual Studio 2008.
> http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
> _______________________________________________
> Dev-cpp-users mailing list
> Dev...@li...
> TO UNSUBSCRIBE: http://www23.brinkster.com/noicys/devcpp/ub.htm
> https://lists.sourceforge.net/lists/listinfo/dev-cpp-users
>

I think what Per refers to is the overhead of initializing and regex
library and calling upon it. What appears as a single function call is
probably several 100 lines of code underneath.

And for small programs it's often better to leave out library dependencies
and reduce compilation times.

Just my 2 cents.

Pranav Negandhi
www.pranavnegandhi.com

>> No regular expressions if he is looking for an exact match. Regular
>> expressions are there to look for patterns. Normal
>> strcmp()/strncmp() is
>> faster when looking for exact matches.
>
> No not at all. You explained an algorithms to look in the middle, then
> depending on less or greater in the lower half or in the upper half etc.
> That needs a lot of lookups if the file is huge.
>
> A regex with the multiline option can find the answer with one line.
> /^WordToFind\s*?(.*)/m
> In the caption(1) you will find the answer, if there is an answer.
>

Why do you talk about scrit? This is the Dev-C++ list, where people
normally uses Dev-C++ to write C or C++ applications. The traditional use
of the word 'script' is for some interpretative programming - often in a
batch file or similar.

Anyway - is the file sorted? Does it contain a huge number of words?

If no to any of the above two questions: Open file. Read one line into a
buffer (for example with fgets()). Find the first separator character
(space, tab or whatever the file is using). Replace the separator with a
'\0'; Compare start of line with the word you want to match. If no match,
load next line and repeat. If match, step past the separator and then eath
any more separator characters. Locate the '\n' at end of the read line,
and replace with a '\0'; Emit the extracted answer.

If the file is huge and sorted, get the file size. Move to the middle of
the file. Read one (possibly partial line). Read the next line. Extract
the word and see if you are above/below the requested language. If you
found the word - extract answer. If word too large, move back to one
quarter of the file and repat. If word was too small, seek to 75% of file
and repeat. All the time remember the <min,max> range that the
expected word must be in. Note that <min> and <max> should represent the
start position of text lines, (or the end-of-file) so when you
specify <min> as seek offset, you do not have to thow away any
potentially partial text lines - you know that it is a full line
and can check that word immediately.

/pwm

On Tue, 3 Jun 2008, Alexsandro Meireles wrote:

> Dear users,
>
> How can I write a script to look a word in a .txt file with two columns (one is the word and the other is the answer I need) and give an answer which is in the second column. In other words, a user types a word, then the script match it with a word in a .txt file (first column) and gives the answer (second column).
>
> Thanks in advance!
>
> Alex.
> _________________________________________________________________
> Cansado de espaço para só 50 fotos? Conheça o Spaces, o site de relacionamentos com até 6,000 fotos!
> http://www.amigosdomessenger.com.br