Accessing elements in a mulimap

Question

I am trying to create a program that takes in data from a .txt or similar file and asks the user for a word to search for. The output should show the keyword in context with the 2 words that were originally in front of it, as well as behind it. (EX: keyword: boy would output "and the boy ran away") I am able to find all instances of the keyword in the file with the equal_range() function, however I do not know how to iterate through the data in the map to access the other words for context. Here is my code so far:

typedef multimap<string, int> templateMap;
templateMap wordMap;
typedef pair<templateMap::iterator, templateMap::iterator> searchTemplate;
searchTemplate search;
typedef pair<templateMap::const_iterator, templateMap::const_iterator> innerIteratorTemplate;
multimap<string, int>::iterator tempMap;
string tempWord;
string keyword;

// omitted code

for (size_t i = 0; !inData.eof(); i++)  
{
    inData >> tempWord;
    wordMap.insert(pair<string, int>(tempWord, i));
}

search = wordMap.equal_range(keyword);

for (multimap<string, int>::iterator itr = search.first; itr != search.second; ++itr)
{
    cout << "The keyword " << keyword << " is found at location " << itr->second << endl;

    tempMap = itr;
    itr->second = itr->second - 2;
    cout << itr->first << endl;
}

I am aware that the code in the for loop at the bottom is wrong, but it was for testing purposes.

Answer 1

You need bidirectional lookup: you need to map a word to its index (this is what wordMap is for) and separately you need to map an index to its word (this is what you're missing). So let's add that, as well as fixing your initial loop:

std::vector<std::string> words;
while (inData >> tempWord) {
    wordMap.insert(std::make_pair(tempWord, words.size()));
    words.push_back(tempWord);
}

Now, we have it bidirectionally - since words allows lookup by index. Thus, we have:

for (auto const& pair : as_range(wordMap.equal_range(keyword))) {
    for (size_t idx = pair.second - 2; idx < pair.second + 3; ++idx) {
        std::cout << words[idx] << ' ';
    }
    std::cout << '\n';
}

as_range() is something that takes a pair of iterators and gives you back something you can use in a range-based for expression. This doesn't account for the bounds of words (if you pick one of the first two or last two words as your keyword), but this should put you on the right track.

---

Additionally, consider using std::map<std::string, std::vector<size_t>> instead of std::multimap<std::string, size_t> if you're always going to be iterating over all the values and don't need iterator stability. See this question for more info.

Answer 2

Given your problem statement, a map is ill-suited since you immediately lose all positional information and you're left trying to find a workaround. If you are willing the hold all your data in a container, you may as well hold it in a vector and perform a linear search. Yes, I know, it will be theoretically slower but there's a fair chance it won't be in practice...

For giggles, here's a completely different approach with <regex> facilities:

// Data.
string const text = "Pack my box with five dozen liquor jugs. The quick brown fox jumps over the lazy dog. The five boxing wizards jump quickly.";

// Word to search for.
string target;
cin >> target;

// Capture the target and up to two words before and after.
regex const context(R"((?:([^\s]+)\s)?(?:([^\s]+)\s)?()" + target + R"()(?:\s([^\s]+))?(?:\s([^\s]+))?)");

// Perform search.
smatch matches;
regex_search(text, matches, context);

// Print results.
copy_if(matches.begin() + 1, matches.end(), ostream_iterator<string>(cout, "\n"), mem_fn(&smatch::value_type::matched));