Parser symbols with different types

Sergei Steshenko · 12-01-2010, 02:29 PM

Quote:

Originally Posted by MTK358

Do you mean macros?

And is there anything wrong with FOLLOWED_BY being a function, or are you just saying it doesn't have to be?

Yes, I mean macros.

I also suggested to "C" GNU dialect - for a reason. Here is a code sample:

Code:

sergei@amdam2:~/junk/gnu_code_blocks> cat -n main.c
     1  #include <stdio.h>
     2
     3  int main()
     4      {
     5      char stream[1000] = "step";
     6      unsigned stream_position = 0;
     7
     8      1 // trivial truth
     9   && ({unsigned result = 0; unsigned saved_position = stream_position;stream_position; if(stream[stream_position] == 's') {stream_position++; result = 1;} fprintf(stderr, "'s' matches\n"); result;})
    10   && ({unsigned result = 0; unsigned saved_position = stream_position;stream_position; if(stream[stream_position] == 't') {stream_position++; result = 1;} fprintf(stderr, "'t' matches\n"); result;})
    11   && ({unsigned result = 0; unsigned saved_position = stream_position;stream_position; if(stream[stream_position] == 'e') {stream_position++; result = 1;} fprintf(stderr, "'e' matches\n"); result;})
    12   && ({unsigned result = 0; unsigned saved_position = stream_position;stream_position; if(stream[stream_position] == 'p') {stream_position++; result = 1;} fprintf(stderr, "'p' matches\n"); result;});
    13
    14      return 0;
    15      }
sergei@amdam2:~/junk/gnu_code_blocks> ~/AFSWD/install/gcc-4.4.5/binsh/gcc main.c -o funny_parser
sergei@amdam2:~/junk/gnu_code_blocks> ./funny_parser
's' matches
't' matches
'e' matches
'p' matches
sergei@amdam2:~/junk/gnu_code_blocks>

- you see, stackable pieces of code, which can be easily converted into macro.

Sergei Steshenko · 12-01-2010, 02:32 PM

Quote:

Originally Posted by MTK358

I'm also starting to think that, and that's what Sergei seems to be leading to.

Sergei is leading towards recursion part after the care is taken of non-recursion part.

MTK358 · 12-01-2010, 02:54 PM

Quote:

Originally Posted by Sergei Steshenko

Yes, I mean macros.

I also suggested to "C" GNU dialect - for a reason. Here is a code sample:

Code:

sergei@amdam2:~/junk/gnu_code_blocks> cat -n main.c
     1  #include <stdio.h>
     2
     3  int main()
     4      {
     5      char stream[1000] = "step";
     6      unsigned stream_position = 0;
     7
     8      1 // trivial truth
     9   && ({unsigned result = 0; unsigned saved_position = stream_position;stream_position; if(stream[stream_position] == 's') {stream_position++; result = 1;} fprintf(stderr, "'s' matches\n"); result;})
    10   && ({unsigned result = 0; unsigned saved_position = stream_position;stream_position; if(stream[stream_position] == 't') {stream_position++; result = 1;} fprintf(stderr, "'t' matches\n"); result;})
    11   && ({unsigned result = 0; unsigned saved_position = stream_position;stream_position; if(stream[stream_position] == 'e') {stream_position++; result = 1;} fprintf(stderr, "'e' matches\n"); result;})
    12   && ({unsigned result = 0; unsigned saved_position = stream_position;stream_position; if(stream[stream_position] == 'p') {stream_position++; result = 1;} fprintf(stderr, "'p' matches\n"); result;});
    13
    14      return 0;
    15      }

What's the advantage of that, seems like using a macro or function would be much less repetetive.

Sergei Steshenko · 12-01-2010, 03:00 PM

Quote:

Originally Posted by MTK358

What's the advantage of that, seems like using a macro or function would be much less repetetive.

As I wrote earlier, first copy-paste in order to understand what can/should be abstracted out/hidden in a macro.

The code I gave shows some important thoughts, for example, 'fprintf' statements are for debugging/tracing; the place where 'result=1' is also a place into which useful code (e.g. adding items to AST) can be added - the code will be execute upon language construct match.

In my http://www.linuxquestions.org/questi...ml#post4177393 post it is prominently written:

Quote:

stackable pieces of code, which can be easily converted into macro.

MTK358 · 12-01-2010, 03:10 PM

Also, I don't get this:

Code:

unsigned saved_position = stream_position;stream_position;

What's the reason for "stream_position;", it looks like it does nothing.

Sergei Steshenko · 12-01-2010, 03:32 PM

Quote:

Originally Posted by MTK358

Also, I don't get this:

Code:

unsigned saved_position = stream_position;stream_position;

What's the reason for "stream_position;", it looks like it does nothing.

A typo - you can safely remove it.

Sergei Steshenko · 12-01-2010, 03:37 PM

By the way, 'fprintf' is in a wrong place too, the code should be:

Code:

sergei@amdam2:~/junk/gnu_code_blocks> cat -n main.c
     1  #include <stdio.h>
     2
     3  int main()
     4      {
     5      char stream[1000] = "step";
     6      unsigned stream_position = 0;
     7
     8      1 // trivial truth
     9   && ({unsigned result = 0; unsigned saved_position = stream_position; if(stream[stream_position] == 's') {stream_position++; result = 1; fprintf(stderr, "'s' matches\n");} result;})
    10   && ({unsigned result = 0; unsigned saved_position = stream_position; if(stream[stream_position] == 't') {stream_position++; result = 1; fprintf(stderr, "'t' matches\n");} result;})
    11   && ({unsigned result = 0; unsigned saved_position = stream_position; if(stream[stream_position] == 'e') {stream_position++; result = 1; fprintf(stderr, "'e' matches\n");} result;})
    12   && ({unsigned result = 0; unsigned saved_position = stream_position; if(stream[stream_position] == 'p') {stream_position++; result = 1; fprintf(stderr, "'p' matches\n");} result;});
    13
    14      return 0;
    15      }
sergei@amdam2:~/junk/gnu_code_blocks>

.

MTK358 · 12-01-2010, 04:04 PM

So that big expression will evalute to true if the input stream at the current point matches "step", otherwise it will return 0 and leave the stream position at the first unsuccessful character, right?

Also, what's the reason for the leading "1 &&"?

Sergei Steshenko · 12-01-2010, 04:25 PM

Quote:

Originally Posted by MTK358

So that big expression will evalute to true if the input stream at the current point matches "step", otherwise it will return 0 and leave the stream position at the first unsuccessful character, right?

Also, what's the reason for the leading "1 &&"?

Not right. The position is advanced with each partial match.

So, if you are to match the whole "step", you need to wrap the code:

Code:

sergei@amdam2:~/junk/gnu_code_blocks> cat -n main.c
     1  #include <stdio.h>
     2
     3  int main()
     4      {
     5      char stream[1000] = "step";
     6      unsigned stream_position = 0;
     7
     8      1 // trivial truth
     9   && (
    10        {
    11        unsigned result = 0;
    12        unsigned saved_position = stream_position;
    13        if(
    14           1
    15        && ({unsigned result = 0; unsigned saved_position = stream_position; if(stream[stream_position] == 's') {stream_position++; result = 1; fprintf(stderr, "'s' matches\n");} result;})
    16        && ({unsigned result = 0; unsigned saved_position = stream_position; if(stream[stream_position] == 't') {stream_position++; result = 1; fprintf(stderr, "'t' matches\n");} result;})
    17        && ({unsigned result = 0; unsigned saved_position = stream_position; if(stream[stream_position] == 'e') {stream_position++; result = 1; fprintf(stderr, "'e' matches\n");} result;})
    18        && ({unsigned result = 0; unsigned saved_position = stream_position; if(stream[stream_position] == 'p') {stream_position++; result = 1; fprintf(stderr, "'p' matches\n");} result;})
    19          )
    20          {
    21          result = 1;
    22            {
    23            fprintf(stderr,"the matching part: ");
    24
    25            unsigned pos;
    26            for(pos = saved_position; pos < stream_position; pos++)
    27              {
    28              fprintf(stderr, "%c", stream[pos]);
    29              }
    30
    31            fprintf(stderr,"\n");
    32            }
    33          }
    34        result;
    35        }
    36      );
    37      return 0;
    38      }
sergei@amdam2:~/junk/gnu_code_blocks> ~/AFSWD/install/gcc-4.4.5/binsh/gcc -Wall -Wextra main.c -o funny_parser
main.c: In function ‘main’:
main.c:15: warning: unused variable ‘saved_position’
main.c:16: warning: unused variable ‘saved_position’
main.c:17: warning: unused variable ‘saved_position’
main.c:18: warning: unused variable ‘saved_position’
main.c:8: warning: value computed is not used
sergei@amdam2:~/junk/gnu_code_blocks> ./funny_parser
's' matches
't' matches
'e' matches
'p' matches
the matching part: step
sergei@amdam2:~/junk/gnu_code_blocks>

The

Code:

1 &&

part is for convenience, i.e. a parser can be formulated as:

Code:

#define TRIVIAL_TRUTH 1
...
TRIVIAL_TRUTH
FOLLOWED_BY(...)
FOLLOWED_BY(...)
...

.

And if backtracking is to be implemented, the outer 'if' used to check matching of the whole 'step' should be accompanied by

Code:

else
  {
  stream_position = saved_position;
  }

, i.e. for cases of lazy writing - say, you first try to match 'step', and then from scratch, when it fails, you want to match 'stop'.

By the way, a good illustration of usefulness of scoping rules - variable names are reused everywhere, and no collision occurs.