Search for a string backwards in a file.

[SK]

Hey folks,

I am searching for a string (say "ABC") backwards in a file.
First I seek to the end.
Then I try to make a check like -

do {
file.clear ();
file.get(c);
file.seekg(-2, std::ios::cur);
} while (c != 'A' && c.readback() != 'B' && c.readback != 'C')
// readback, hypotheticl func, problem here)
Is there some function/trick like peek that can instead read backwards
one character at a time?
Or is there a better way to solve this problem.

Thank you.

[John Harrison]

"SK" <sk******@redif fmail.com> wrote in message
news:83******** *************** ***@posting.goo gle.com...

Hey folks,

I am searching for a string (say "ABC") backwards in a file.
First I seek to the end.
Then I try to make a check like -

do {
file.clear ();
file.get(c);
file.seekg(-2, std::ios::cur);
} while (c != 'A' && c.readback() != 'B' && c.readback != 'C')
// readback, hypotheticl func, problem here)
Is there some function/trick like peek that can instead read backwards
one character at a time?
Or is there a better way to solve this problem.

Thank you.

Is the file small enough to read the whole file into memory? If so then read
the whole file into a string and search in the string not in the file.
Anything else rapidly gets very complicated and also not terribly efficient.
Files aren't designed to be read backwards, I would prefer to redesign your
file format so that you don't need to read backwards than to actually
attempt this.

And no there is no quick trick to do this.

john

[Karl Heinz Buchegger]

SK wrote:

Hey folks,

I am searching for a string (say "ABC") backwards in a file.
First I seek to the end.
Then I try to make a check like -

do {
file.clear ();
file.get(c);
file.seekg(-2, std::ios::cur);
} while (c != 'A' && c.readback() != 'B' && c.readback != 'C')
// readback, hypotheticl func, problem here)
Is there some function/trick like peek that can instead read backwards
one character at a time?
No.
Or is there a better way to solve this problem.

Seek to the end.

Seek back a number of characters
Read a number of characters into a buffer.
Search that buffer from the end.

If you find the pattern -> fine
If you don't find the pattern: seek back a number
of characters into the buffer, check the buffer
and repeat.
--
Karl Heinz Buchegger
kb******@gascad .at

[John Harrison]

"Karl Heinz Buchegger" <kb******@gasca d.at> wrote in message
news:40******** *******@gascad. at...

SK wrote:

Hey folks,

I am searching for a string (say "ABC") backwards in a file.
First I seek to the end.
Then I try to make a check like -

do {
file.clear ();
file.get(c);
file.seekg(-2, std::ios::cur);
} while (c != 'A' && c.readback() != 'B' && c.readback != 'C')
// readback, hypotheticl func, problem here)
Is there some function/trick like peek that can instead read backwards
one character at a time?

No.

Or is there a better way to solve this problem.

Seek to the end.

Seek back a number of characters
Read a number of characters into a buffer.
Search that buffer from the end.

If you find the pattern -> fine
If you don't find the pattern: seek back a number
of characters into the buffer, check the buffer
and repeat.

And don't forget to deal with the case where the string you are searching
for straddles one of your seek positions. I.e. if you aren't careful, one
half of the string ends up in one buffer and the other half in another
buffer, so you never find it.

john

[Karl Heinz Buchegger]

John Harrison wrote:

"Karl Heinz Buchegger" <kb******@gasca d.at> wrote in message
news:40******** *******@gascad. at...

SK wrote:

Hey folks,

I am searching for a string (say "ABC") backwards in a file.
First I seek to the end.
Then I try to make a check like -

do {
file.clear ();
file.get(c);
file.seekg(-2, std::ios::cur);
} while (c != 'A' && c.readback() != 'B' && c.readback != 'C')
// readback, hypotheticl func, problem here)
Is there some function/trick like peek that can instead read backwards
one character at a time?

No.

Or is there a better way to solve this problem.

Seek to the end.

Seek back a number of characters
Read a number of characters into a buffer.
Search that buffer from the end.

If you find the pattern -> fine
If you don't find the pattern: seek back a number
of characters into the buffer, check the buffer
and repeat.

And don't forget to deal with the case where the string you are searching
for straddles one of your seek positions. I.e. if you aren't careful, one
half of the string ends up in one buffer and the other half in another
buffer, so you never find it.

:-)
It's a tricky thing which could be solved with letting the reads overlap:

+---------------------------+
+---------------------------+

| |
Overlapping area, large enough that the pattern will fit
into it.

But there should be left something to think about for the OP.

The best thing the OP could do is: Avoid that topic at all by redesigning
the file format.

--
Karl Heinz Buchegger
kb******@gascad .at

[Henrik Vallgren]

How about reversing "ABC" and doing a plain search?

Henrik Vallgren

"SK" <sk******@redif fmail.com> skrev i meddelandet
news:83******** *************** ***@posting.goo gle.com...

Hey folks,

I am searching for a string (say "ABC") backwards in a file.
First I seek to the end.
Then I try to make a check like -

do {
file.clear ();
file.get(c);
file.seekg(-2, std::ios::cur);
} while (c != 'A' && c.readback() != 'B' && c.readback != 'C')
// readback, hypotheticl func, problem here)
Is there some function/trick like peek that can instead read backwards
one character at a time?
Or is there a better way to solve this problem.

Thank you.

[SK]

"John Harrison" <jo************ *@hotmail.com> wrote in message news:<2j******* *****@uni-berlin.de>...

"SK" <sk******@redif fmail.com> wrote in message
news:83******** *************** ***@posting.goo gle.com...

Hey folks,

I am searching for a string (say "ABC") backwards in a file.
First I seek to the end.
Then I try to make a check like -

do {
file.clear ();
file.get(c);
file.seekg(-2, std::ios::cur);
} while (c != 'A' && c.readback() != 'B' && c.readback != 'C')
// readback, hypotheticl func, problem here)
Is there some function/trick like peek that can instead read backwards
one character at a time?
Or is there a better way to solve this problem.

Thank you.
Is the file small enough to read the whole file into memory?

No, it goes in several mbs.
If so then read the whole file into a string and search in the string not in the file.
Anything else rapidly gets very complicated and also not terribly efficient.
Files aren't designed to be read backwards, I would prefer to redesign your
file format so that you don't need to read backwards than to actually
attempt this.

Wish I could redesign the file format, but can't actually :-(
But what I am parsing are timestamps in a file. I need to know the
first and the last timestamp in a given file.
And no there is no quick trick to do this.

Thanks.

[John Harrison]

"SK" <sk******@redif fmail.com> wrote in message
news:83******** *************** ***@posting.goo gle.com...

"John Harrison" <jo************ *@hotmail.com> wrote in message

news:<2j******* *****@uni-berlin.de>...

"SK" <sk******@redif fmail.com> wrote in message
news:83******** *************** ***@posting.goo gle.com...

Hey folks,

I am searching for a string (say "ABC") backwards in a file.
First I seek to the end.
Then I try to make a check like -

do {
file.clear ();
file.get(c);
file.seekg(-2, std::ios::cur);
} while (c != 'A' && c.readback() != 'B' && c.readback != 'C')
// readback, hypotheticl func, problem here)
Is there some function/trick like peek that can instead read backwards
one character at a time?
Or is there a better way to solve this problem.

Thank you.

Is the file small enough to read the whole file into memory?

No, it goes in several mbs.
If so then read

the whole file into a string and search in the string not in the file.
Anything else rapidly gets very complicated and also not terribly efficient. Files aren't designed to be read backwards, I would prefer to redesign your file format so that you don't need to read backwards than to actually
attempt this.

Wish I could redesign the file format, but can't actually :-(
But what I am parsing are timestamps in a file. I need to know the
first and the last timestamp in a given file.

Then Karl's suggestion is the right one, read blocks of data from the end of
a file, and search backwards though each block, repeat if you don't find
anything, and watch out for the timestamp falling half way between two
blocks. In practise this means that the blocks have to overlap sufficiently
so that the timestamp will always be contained entirely within one block.

john

[Siemel Naran]

"SK" <sk******@redif fmail.com> wrote in message

I am searching for a string (say "ABC") backwards in a file.
First I seek to the end.
Then I try to make a check like -

do {
file.clear ();
file.get(c);
file.seekg(-2, std::ios::cur);
} while (c != 'A' && c.readback() != 'B' && c.readback != 'C')
// readback, hypotheticl func, problem here)
Is there some function/trick like peek that can instead read backwards
one character at a time?
Or is there a better way to solve this problem.
This approach seems fine to me. Others suggested reading the file into
blocks of strings. But I'm not sure if there's a need to do this as the
default fstreams already read the file into blocks of strings (though the
standard requires them to, but they do for file streams). As for how they
normally handle reading from the end of the file, I don't know for sure, but
I think it's like this: when the user reads the last character, then read
the last N chars into memory where N is the size of a block, position the
streambuf's get pointer to the last character, return the character at the
get pointer.
file.get(c);
file.seekg(-2, std::ios::cur);

The above might be more clearly expressed with

c = file.peek();
file.seekg(-1, std::ios::cur);

Also, since you're looking for a string of length 3, you could save the last
3 chars read into variables like c1, c2, c3. Or even an array, vector, etc.
Then even if you encounter an 'A' you could ignore it if the next characters
are not 'BC'.

Finally, to get improved performance you could work directly with the
underlying streambuf. Call file.rdbuf() to get a pointer to the streambuf.
It will really be a filebuf, but pretend you don't know this. To position
the stream to the beginning or end use pubseekpos. To position the stream
relative to the current position use pubseekoff. To get the current
character as in peek use sgetc.

If this is not fast enough, you could even write your own class derived from
filebuf that implements a sbumpcminus() function that gets the current
character then decrements the get pointer. Use setg to set the get pointers
and range of the get area. But this is getting really advanced.

peek at the current character

[Jorge Rivera]

SK wrote:

Hey folks,

I am searching for a string (say "ABC") backwards in a file.
First I seek to the end.
Then I try to make a check like -

do {
file.clear ();
file.get(c);
file.seekg(-2, std::ios::cur);
} while (c != 'A' && c.readback() != 'B' && c.readback != 'C')
// readback, hypotheticl func, problem here)
Is there some function/trick like peek that can instead read backwards
one character at a time?
Or is there a better way to solve this problem.

Thank you.

You can do something like this.

std::string toFind = "ABC";
std::string reversed;
std::copy(toFin d.rbegin(), toFind.rend(),s td::front_inser ter(reversed));

std::ifstream ifs("MyFile.txt ", std::ios::in);
if(ifs)
{
std::string fileContent = std::string(
std::istreambuf _iterator<char> (ifs),
std::istreambuf _iterator<char> ());

std::string::si ze_type found_position = fileContent .rfind(reversed );
if(std::string: :npos != found_position)
std::cout<<"Req uested string found at position:
"<<found_positi ons<<std::endl;
}

(Forgive any typos...)

JLR