Thread: what would I use to count number of words

Frankly, I fear you've lead this thread astray.

You don't need a "Finite State Machine" or convoluted logic to count words. Certainly, no goto's are required (although I use them w/o guilt, as K&R advocated in the White Book).

To count words, I just:

Code:

add_word = 1

while char's remain unread in the string
  read a char in the string.
  
  If it's a letter, and add_word flag is 1 
    increment the word count, 
    set add_word flag to 0

  if char is a space, 
    set add_word flag to 1

loop

This is without actual coding & testing, but I believe it will do the job.

The add_word variable is key to a simple loop logic for this.

Adak

Originally Posted by Adak

Code:

while char's remain unread in the string
  read a char in the string.
  
  If it's a letter, and add_word flag is 1 
    increment the word count, 
    set add_word flag to 0

  if char is a space, 
    set add_word flag to 1

Pretty much how I'd do it. You might add that the initial state of add_word depends on the first character in the string. I gave a state machine example because it came up as a potential method. And the recursive version was pretty much just showing off

Originally Posted by Salem

Code:

typedef enum {
    inWord,
    inSpace
} state_t;

I missed this originally. According to the standard, all type names ending in _t are reserved. I'm just picking on you

Originally Posted by brewbuck

As I said, unless the variable was passed from another function, t
he compiler can determine whether this is the case.



So long as that other function is in the same source module, the compiler has no trouble figuring it out. Some compilers, like Intel's ICC compiler, can even perform such analysis across different translation units (.c files).



For sufficiently complex code, it is far more important for the code to be comprehensible than it is to be 100% "optimal." I've never encountered a case where goto caused previously unacceptably slow code to become acceptable. On the other hand, I have, many times, made simple algorithmic changes which sped up code by factors of 100 or more. And I've got several million lines of code under my belt.



Maintain it all you want. Whether you're correct or not, most people disagree. Being right isn't going to help you when you get into an argument. People get religious about this stuff.

Perhaps for the type of code that you need to write it's fine. But for what I do, efficiency is paramount. For real time programming, state machines are usually implemented via look up tables or using the dreaded goto. I'll bet there is a reason that Ragel's state machine compile produces C code that employs gotos!
http://www.cs.queensu.ca/~thurston/ragel/

Gotos need not be harder to read than using switch, provided that they are used rationally. If you had a document that contained the state diagram for a FSM and then had well documented C code that implemented the state machine using gotos it should be no harder to understand than using the switch statement, which actually is an extra, unneeded layer of "cruft" for this application IMHO.

Originally Posted by SevenThunders

I'll bet there is a reason that Ragel's state machine compile produces C code that employs gotos!

The reason is that Ragel produces computer-generated code. It's just easier to do it that way. (And yes, I've had this discussion with Ragel's author).

Gotos need not be harder to read than using switch, provided that they are used rationally. If you had a document that contained the state diagram for a FSM and then had well documented C code that implemented the state machine using gotos it should be no harder to understand than using the switch statement, which actually is an extra, unneeded layer of "cruft" for this application IMHO.

I'm going to end my participation in this discussion at this point. Write code as you see fit. This is tedious.

Originally Posted by Adak

Frankly, I fear you've lead this thread astray.

You don't need a "Finite State Machine" or convoluted logic to count words. Certainly, no goto's are required (although I use them w/o guilt, as K&R advocated in the White Book).

To count words, I just:

Code:

add_word = 1

while char's remain unread in the string
  read a char in the string.
  
  If it's a letter, and add_word flag is 1 
    increment the word count, 
    set add_word flag to 0

  if char is a space, 
    set add_word flag to 1

loop

This is without actual coding & testing, but I believe it will do the job.

The add_word variable is key to a simple loop logic for this.

Adak

Uh huh. Your state variable is add_word_flag. It's entirely equivalent to brewbuck's Mealy state machine. I happen to think that these kinds of problems, which is a kind of recognizer, are most easily thought of in terms of state machines. What state are we in when we want to increment? What state are we in when we are waiting for a new word, etc. etc. But maybe that's just the way my brain works.

Originally Posted by Salem

Which of these refers to names ending in _t ?

Perhaps I am entirely hallucinating this requirement. As Mr. Sinkula said, maybe it's from POSIX, not standard C. If so, I've got a few butts to kick, mainly my own.

hey buddy u just need to count da words, so here is a bit of code, m naive here but just tried my best to put out dis 4 u man

Code:

char a[100];
int count=0;
printf("Enter the sentence:");
gets(a);
for(i=0;a[i]!='\0';i++)
{
if(a[i]==32)  //checks if it is space
{
if(i!=strlen(a))     //  it must not be the space after the last word
{
if(a[i]==a[i-1])   //checks if two of more spaces are in sequence
count++;
}
}
}

work with dis n give me feedback buddy