I have a string (nested strings even) that are formatted like a C++ braced initializer list. I want to tokenize them one level at a time into a vector of strings.
So when I input "{one, two, three}"
to the function should output a three element vector
"one"
,
"two"
,
"three"
To complicate this, it needs to support quoted tokens and preserve nested lists:
Input String: "{one, {2, \"three four\"}}, \"five, six\", {\"seven, eight\"}}"
Output is a four element vector:
"one"
,
"{2, \"three four\"}"
,
"five, six"
,
"{\"seven, eight\"}"
I've looked at a few other SO posts:
Using Boost Tokenizer escaped_list_separator with different parameters
Boost split not traversing inside of parenthesis or braces
And used those to start a solution, but this seems slightly too complicated for the tokenizer (because of the braces):
#include <boost/algorithm/string.hpp>
#include <boost/tokenizer.hpp>
std::vector<std::string> TokenizeBracedList(const std::string& x)
{
std::vector<std::string> tokens;
std::string separator1("");
std::string separator2(",\n\t\r");
std::string separator3("\"\'");
boost::escaped_list_separator<char> elements(separator1, separator2, separator3);
boost::tokenizer<boost::escaped_list_separator<char>> tokenizer(x, elements);
for(auto i = std::begin(tokenizer); i != std::end(tokenizer); ++i)
{
auto token = *i;
boost::algorithm::trim(token);
tokens.push_back(token);
}
return tokens;
}
With this, even in the trivial case, it doesn't strip the opening and closing braces.
Boost and C++17 are fair game for a solution.
Simple (Flat) Take
Defining a flat data structure like:
We can define an X3 parser like:
Live On Wandbox
Prints
Nested Data
Changing the datastructure to be a bit more specific/realistic:
Now we can change the grammar, as we no longer need to treat
sublist
as if it is a string:Everything "still works": Live On Wandbox
Prints: