在二进制数据中搜索文本

Question

I've a binary data which contains a text. The text is known. What could be a fast method to search for that text:

As an eg.

This is text 1---
!@##$%%#^%&!%^$! <= Assume this line is 3 MB of binary data
Now, This is text 2 ---
!@##$%%#^%&!%^$! <= Assume this line is 2.5 MB of binary data
This is text 3 ---

How can I search for text This is text 2 .

Currently I'm doing like:

size_t count = 0;
size_t s_len = strlen("This is text 2");

//Assume data_len is length of the data from which text is to be found and data is pointer (char*) to the start of it.
for(; count < data_len; ++count)
{
    if(!memcmp("This is text 2", data + count, s_len)
    {
         printf("%s\n", "Hurray found you...");
    }
}

• Is there any other way, more efficient way to do this
• Will replacing ++count logic with memchr('T') logic help <= Please ignore if this statement is not clear
• what should be the average case big-O comlexity of memchr

Answer 1

There are algorithms for doing exactly this with better complexity than repeated memcmp (which is implemented the obvious way and has the obvious complexity for near matches).

Famous algorithms are Boyer-Moore and Knuth-Morris-Pratt . These are only two examples. The general category in which these fall is "string matching".

Answer 2

There's nothing in standard C to help you, but there is a GNU extension memmem() that does this:

#define TEXT2 "This is text 2"

char *pos = memmem(data, data_len, TEXT2, sizeof(TEXT2));

if (pos != NULL)
    /* Found it. */

If you need to be portable to systems that don't have this, you could take the glibc implementation of memmem() and incorporate it into your program.

Answer 3

I know that the question is about C programming language, but have you tried to use strings unix tool: http://en.wikipedia.org/wiki/Strings_(Unix ) with grep ?

$ strings datafile | grep "your text"

EDIT:

If you want to use C, I suggest to do this simple optimization:

size_t count = 0;
size_t s_len = strlen("This is text 2");

for(; count < data_len; ++count)
{
    if (!isprint(data[count])) continue;

    if(!memcmp("This is text 2", data + count, s_len)
    {
     printf("%s\n", "Hurray found you...");
    }
}

If you want a better performance, I suggest you to search and use a string matching algorithm.