Splitting a string into an array words

S

Simon

Well, the title's pretty descriptive; how would I be able to take a
line of input like this:

getline(cin,mostrecentline);

And split into an (flexible) array of strings. For example: "do this
action"
would go to:

item 0: do
item 1: this
item 2: action

Thanks in advance,
Simon
 
M

Marcus Kwok

Simon said:
Well, the title's pretty descriptive; how would I be able to take a
line of input like this:

getline(cin,mostrecentline);

And split into an (flexible) array of strings. For example: "do this
action"
would go to:

item 0: do
item 1: this
item 2: action

If you are splitting the words by whitespace, you could create a
std::istringstream and push them into a std::vector<std::string>.

Something like: (untested and uncompiled)

std::istringstream line(mostrecentline);
std::vector<std::string> words;
std::string temp;

while (line >> temp) {
words.push_back(temp);
}

You will need to #include <sstream>, <string>, and <vector> for this
method.
 
D

Daniel T.

"Simon said:
Well, the title's pretty descriptive; how would I be able to take a
line of input like this:

getline(cin,mostrecentline);

And split into an (flexible) array of strings. For example: "do this
action"
would go to:

item 0: do
item 1: this
item 2: action

#include <vector>
#include <string>
#include <iostream>
#include <iterator>
// other includes as necessary

template < typename OutIt >
void split( const std::string& in, OutIt result )
{
// add code here...
}

int main() {
string seed = "step1";
vector<string> result;
split( seed, back_inserter( result ) );
assert( result.size() == 1 );
assert( result[0] == "step1" );
std::cout << "You did it! Good job!\n"
}

Run the above program. Make chances to the part labeled "add code here"
until the program compiles and prints out "You did it! Good job!".

When it does, post back here with the code and I'll help you with the
next step.
 
R

Rolf Magnus

Simon said:
Well, the title's pretty descriptive; how would I be able to take a
line of input like this:

getline(cin,mostrecentline);

And split into an (flexible) array of strings.

What do you mean by "flexible", and which separators do you want to use?
For example: "do this
action"
would go to:

item 0: do
item 1: this
item 2: action

In this case, I'd use a stringstream and operator>>.
 
M

Mark P

Simon said:
Well, the title's pretty descriptive; how would I be able to take a
line of input like this:

getline(cin,mostrecentline);

And split into an (flexible) array of strings. For example: "do this
action"
would go to:

item 0: do
item 1: this
item 2: action

Thanks in advance,
Simon

Here's a little tokenizer fcn I've used before. Not necessarily the
most elegant or compact way to do this (and criticisms are welcomed):

// Populates "out" with delimited substrings of "in".
int tokenize (const string& in, vector<string>& out, const char* delims)
{
string::size_type wordStart = 0; // current word start position
string::size_type wordEnd = 0; // last word end position

while (true)
{
wordStart = in.find_first_not_of(delims,wordEnd);
if (wordStart == in.npos)
break;
wordEnd = in.find_first_of(delims,wordStart);
if (wordEnd == in.npos)
wordEnd = in.size();
out.push_back(in.substr(wordStart,wordEnd - wordStart));
}
return out.size();
}

Mark
 
D

Daniel T.

Mark P said:
Here's a little tokenizer fcn I've used before. Not necessarily the
most elegant or compact way to do this (and criticisms are welcomed):

Well since criticisms are welcomed... :)
// Populates "out" with delimited substrings of "in".
int tokenize (const string& in, vector<string>& out, const char* delims)
{
string::size_type wordStart = 0; // current word start position
string::size_type wordEnd = 0; // last word end position

while (true)
{
wordStart = in.find_first_not_of(delims,wordEnd);
if (wordStart == in.npos)
break;
wordEnd = in.find_first_of(delims,wordStart);
if (wordEnd == in.npos)
wordEnd = in.size();
out.push_back(in.substr(wordStart,wordEnd - wordStart));
}
return out.size();
}

From least important to most important:

1) The while true and break is not a style I prefer.

2) Returning out.size() isn't very useful since the caller can find out
what out.size() equals without the functions help.

3) It only works for vectors, I'd write something that works for deques
and lists as well.

4) A cyclomatic complexity of 4 seems a tad excessive for what is
supposed to be such a simple job. You can drop that to 3 by removing
the unnecessary "if (wordEnd == in.npos)" logic. Heeding item (1)
above can reduce the complexity to 2.

Here's how I would write it:

template <typename OutIt>
void tokenize( const string& str, OutIt os, const string& delims = " ")
{
string::size_type start = str.find_first_not_of( delims );
while ( start != string::npos ) {
string::size_type end = str.find_first_of( delims, start );
*os++ = str.substr( start, end - start );
start = str.find_first_not_of( delims, end );
}
}
 
M

Mark P

Daniel said:
Well since criticisms are welcomed... :)


From least important to most important:

1) The while true and break is not a style I prefer.

Fair enough-- I'm not a fan either, but see my comment to item 4.
2) Returning out.size() isn't very useful since the caller can find out
what out.size() equals without the functions help.

True. In my case, I pulled this function out of some actual code where
the return value is sometimes used as a check. E.g., when parsing a
particular file format, I expect a certain number of tokens per line.
It saves the calling function a line of code by having the size of out
returned automatically (and of course this fcn is called in multiple
places).
3) It only works for vectors, I'd write something that works for deques
and lists as well.

Agreed, I very much prefer your templated approach that takes any Output
Iterator. In my case, using a known type allowed me to return the
container size (cf. item 2), but this is just my own particular
situation and at times excessive code parsimony.
4) A cyclomatic complexity of 4 seems a tad excessive for what is
supposed to be such a simple job. You can drop that to 3 by removing
the unnecessary "if (wordEnd == in.npos)" logic. Heeding item (1)
above can reduce the complexity to 2.

Here's how I would write it:

template <typename OutIt>
void tokenize( const string& str, OutIt os, const string& delims = " ")
{
string::size_type start = str.find_first_not_of( delims );
while ( start != string::npos ) {
string::size_type end = str.find_first_of( delims, start );
*os++ = str.substr( start, end - start );
start = str.find_first_not_of( delims, end );
}
}

Looks good. In my case it was a bit more complicated because I also
have an additional parameter for a comment character. When a comment
character is encountered at the beginning of a token, that token is
discarded and the loop breaks. (So in my original implementation there
were multiple breakpoints out of the loop, although I hastily trimmed
these before I posted my code, thereby leaving some unattractive vestiges.)

In any event, I appreciate your comments and don't mean to simply make
excuses and argue all of your points. The only significant hitch to my
adopting your cleaner implementation is that I really do need support
for the comment character break. Luckily this is just a bit of a little
file parser I use for testing, so I don't stress too much about these
details, but feel free to propose a svelte implementation that supports
a comment char. :)

Mark
 
D

Daniel T.

template <typename OutIt>
void tokenize( const string& str, OutIt os, const string& delims = " ")
{
string::size_type start = str.find_first_not_of( delims );
while ( start != string::npos ) {
string::size_type end = str.find_first_of( delims, start );
*os++ = str.substr( start, end - start );
start = str.find_first_not_of( delims, end );
}
}

Looks good. In my case it was a bit more complicated because I also
have an additional parameter for a comment character. When a comment
character is encountered at the beginning of a token, that token is
discarded and the loop breaks. (So in my original implementation there
were multiple breakpoints out of the loop, although I hastily trimmed
these before I posted my code, thereby leaving some unattractive vestiges.)

In any event, I appreciate your comments and don't mean to simply make
excuses and argue all of your points. [/QUOTE]

No problem. Your code was rather good in general, I only saw a few nits
to pick at.
The only significant hitch to my
adopting your cleaner implementation is that I really do need support
for the comment character break. Luckily this is just a bit of a little
file parser I use for testing, so I don't stress too much about these
details, but feel free to propose a svelte implementation that supports
a comment char. :)

If I understand what you mean then:

void tokenize( const string& str, OutIt os, const string& delims = " ",
char comment = '\0' )
{
string::size_type start = str.find_first_not_of( delims );
while ( start != string::npos && start[0] != comment ) {
string::size_type end = str.find_first_of( delims, start );
*os++ = str.substr( start, end - start );
start = str.find_first_not_of( delims, end );
}
}

Of course you should probably change the defaults to whatever is most
common in your code...
 
D

Daniel T.

2) Returning out.size() isn't very useful since the caller can find out
what out.size() equals without the functions help.

True. In my case, I pulled this function out of some actual code where
the return value is sometimes used as a check. E.g., when parsing a
particular file format, I expect a certain number of tokens per line.
It saves the calling function a line of code by having the size of out
returned automatically (and of course this fcn is called in multiple
places).[/QUOTE]

Here you go, now it returns the size. :)

int tokenize( const string& str, OutIt os, const string& delims = " ",
char comment = '\0' )
{
int result = 0;
string::size_type start = str.find_first_not_of( delims );
while ( start != string::npos && start[0] != comment ) {
string::size_type end = str.find_first_of( delims, start );
*os++ = str.substr( start, end - start );
++result;
start = str.find_first_not_of( delims, end );
}
return result;
}
 
A

Alex Vinokur

Simon said:
Well, the title's pretty descriptive; how would I be able to take a
line of input like this:

getline(cin,mostrecentline);

And split into an (flexible) array of strings.
[snip]

See "Splitting string into vector of vectors":
http://groups.google.com/group/sources/msg/77993fb8841382c8
http://groups.google.com/group/perfo/msg/9d49a1be3a5c6335
http://groups.google.com/group/perfo/msg/f3c775cf7e3cdcf0


Alex Vinokur
email: alex DOT vinokur AT gmail DOT com
http://mathforum.org/library/view/10978.html
http://sourceforge.net/users/alexvn
 
O

Old Wolf

Daniel said:
int tokenize( const string& str, OutIt os, const string& delims = " ",
char comment = '\0' )

You should return the size type of the output iterator,
rather than int.

I am suspicious of the code. Suppose str is "x".
{
int result = 0;
string::size_type start = str.find_first_not_of( delims );

start == 0.
while ( start != string::npos && start[0] != comment ) {

condition is true
string::size_type end = str.find_first_of( delims, start );

end is string::npos
*os++ = str.substr( start, end - start );

Here you subtract a value from npos. I am not sure if this is a
legal operation (although it will happen to work on my system).
++result;
start = str.find_first_not_of( delims, end );

is npos a legal argument for the second parameter to find_first_not_of
?
 
D

Daniel T.

"Old Wolf said:
You should return the size type of the output iterator,
rather than int.

That would be fine too...

I had to fix the code, start[0] of course is silly.

template < typename OutIt >
int tokenize( const string& str, OutIt os, const string& delims = " ",
char comment = '\0' )
{
int result = 0;
string::size_type start = str.find_first_not_of( delims );
while ( start != string::npos && str[start] != comment ) {
string::size_type end = str.find_first_of( delims, start );
*os++ = str.substr( start, end - start );
++result;
start = str.find_first_not_of( delims, end );
}
return result;
}

end is string::npos


Here you subtract a value from npos. I am not sure if this is a
legal operation (although it will happen to work on my system).

I have several sources that say that npos is "The largest possible value
of type size_type." Most importantly, it is *not* a flag but a defined
value. Subtracting from a value is quite legal.

is npos a legal argument for the second parameter to find_first_not_of

Or the broader question, what is the defined result of find_first_not_of
if the second argument is greater than str.length().

It could be that my implementation (and yours) is doing the wrong thing,
Stroustrup in "The C++ Programming Language" says that in general
specifying a index >= the length() should throw an exception which would
require me to add another conditional here:

start = ( end == string::npos ) ?
end : str.find_first_not_of( delims, end );

Maybe someone can check the standard for me?
 
D

davidrubin

Mark said:
Here's a little tokenizer fcn I've used before. Not necessarily the
most elegant or compact way to do this (and criticisms are welcomed):

// Populates "out" with delimited substrings of "in".
int tokenize (const string& in, vector<string>& out, const char* delims)
{
string::size_type wordStart = 0; // current word start position
string::size_type wordEnd = 0; // last word end position

while (true)
{
wordStart = in.find_first_not_of(delims,wordEnd);
if (wordStart == in.npos)
break;
wordEnd = in.find_first_of(delims,wordStart);
if (wordEnd == in.npos)
wordEnd = in.size();
out.push_back(in.substr(wordStart,wordEnd - wordStart));
}
return out.size();
}

Mark

Along the same lines, here is something from a while back...

http://groups.google.com/group/comp...st&q=davidrubin+split&rnum=1#79258d2ea71e3e03
 

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments. After that, you can post your question and our members will help you out.

Ask a Question

Members online

No members online now.

Forum statistics

Threads
473,744
Messages
2,569,482
Members
44,901
Latest member
Noble71S45

Latest Threads

Top