I'm creating an index file in JSON, which I'm using as a sort-of-database index for a javascript application I'm working on.

My index will look like this:

{
    "_id": "aplex_indices.json",
    "indexAB": {
        "title": {
            "Shawshank Redemption": [
                "0"
            ],
            "Godfather": [
                "1"
            ],
            "Godfather 2": [
                "2"
            ],
            "Pulp Fiction": [
                "3"
            ],
            "The Good, The Bad and The Ugly": [
                "4"
            ],
            "12 Angry Men": [
                "5"
            ],
            "The Dark Knight": [
                "6"
            ],
            "Schindlers List": [
                "7"
            ],
            "Lord of the Rings - Return of the King": [
                "8"
            ],
            "Fight Club": [
                "9"
            ],
            "Star Wars Episode V": [
                "10"
            ],
            "Lord Of the Rings - Fellowship of the Ring": [
                "11"
            ],
            "One flew over the Cuckoo's Nest": [
                "12"
            ],
            "Inception": [
                "13"
            ],
            "Godfellas": [
                "14"
            ]
        },
        "year": {
            "1994": [
                "0",
                "3"
            ],
            "1972": [
                "1"
            ],
            "1974": [
                "2"
            ],
            "1966": [
                "4"
            ],
            "1957": [
                "5"
            ],
            "2008": [
                "6"
            ],
            "1993": [
                "7"
            ],
            "2003": [
                "8"
            ],
            "1999": [
                "9"
            ],
            "1980": [
                "10"
            ],
            "2001": [
                "11"
            ],
            "1975": [
                "12"
            ],
            "2010": [
                "13"
            ],
            "1990": [
                "14"
            ]
        }
    }
}

So for every keyword (like Pulp Fiction), I'm storing the matching document-id(s).

My problem is with integers/numbers/non-string data, like the release year in the above example. This is stored as a string, while I had hoped it would be stored as a number.

I'm creating the index entries like this:

// indices = the current index file
// doc = the document to update the index with
// priv.indices = all indices defined for this application instance
// priv.indices.fields = index fields e.g. "year", "director", "title"
// priv.indices.name = name of this index

priv.updateIndices = function (indices, doc) {
var i, j, index, value, label, key, l = priv.indices.length;

// loop all indices to add document
for (i = 0; i < l; i += 1) {
  index = {};
  index.reference = priv.indices[i];
  index.reference_size = index.reference.fields.length;
  index.current = indices[index.reference.name];

  for (j = 0; j < index.reference_size; j += 1) {
    label = index.reference.fields[j];    // like "year"
    value = doc[label];                   // like 1985

    // if document has a label field (e.g. doc.year = 1985)
    if (value !== undefined) {

      // check if the index file already contains an entry for 1985
      index.current_size = priv.getObjectSize(index.current[label]);
      
      if (index.current_size > 0) {
        // check if the document id is already in the index
        // in case the data is updated (e.g. change 1982 to 1985)
        key = priv.searchIndexByValue(
          index.current[label],
          doc._id,
          "key"
        );
        if (!!key) {
          delete index.current[label][key];
        }
      }
      // create a new array if 1985 is not in the index yet
      if (index.current[label][value] === undefined) {
        index.current[label][value] = [];
      }
      // add the document id to an existing entry
      index.current[label][value].push(doc._id);
    }
  }
}
return indices;
};

This works fine, except that fields I want to store as non-strings (integers, numbers or datetime), like the year in the above example end up as strings in my index.

Question:
Is it at all possible to store "non-string" types in a JSON document? If so, can I also store the key of a key/value pair as a "non-string" element.

If not, would I have to add a parameter to my index definitions declaring the type of each key in order to modify the key-string when I run into it or is there a better way to do it?

Thanks!

I'm creating an index file in JSON, which I'm using as a sort-of-database index for a javascript application I'm working on.

My index will look like this:

{
    "_id": "aplex_indices.json",
    "indexAB": {
        "title": {
            "Shawshank Redemption": [
                "0"
            ],
            "Godfather": [
                "1"
            ],
            "Godfather 2": [
                "2"
            ],
            "Pulp Fiction": [
                "3"
            ],
            "The Good, The Bad and The Ugly": [
                "4"
            ],
            "12 Angry Men": [
                "5"
            ],
            "The Dark Knight": [
                "6"
            ],
            "Schindlers List": [
                "7"
            ],
            "Lord of the Rings - Return of the King": [
                "8"
            ],
            "Fight Club": [
                "9"
            ],
            "Star Wars Episode V": [
                "10"
            ],
            "Lord Of the Rings - Fellowship of the Ring": [
                "11"
            ],
            "One flew over the Cuckoo's Nest": [
                "12"
            ],
            "Inception": [
                "13"
            ],
            "Godfellas": [
                "14"
            ]
        },
        "year": {
            "1994": [
                "0",
                "3"
            ],
            "1972": [
                "1"
            ],
            "1974": [
                "2"
            ],
            "1966": [
                "4"
            ],
            "1957": [
                "5"
            ],
            "2008": [
                "6"
            ],
            "1993": [
                "7"
            ],
            "2003": [
                "8"
            ],
            "1999": [
                "9"
            ],
            "1980": [
                "10"
            ],
            "2001": [
                "11"
            ],
            "1975": [
                "12"
            ],
            "2010": [
                "13"
            ],
            "1990": [
                "14"
            ]
        }
    }
}

So for every keyword (like Pulp Fiction), I'm storing the matching document-id(s).

My problem is with integers/numbers/non-string data, like the release year in the above example. This is stored as a string, while I had hoped it would be stored as a number.

I'm creating the index entries like this:

// indices = the current index file
// doc = the document to update the index with
// priv.indices = all indices defined for this application instance
// priv.indices.fields = index fields e.g. "year", "director", "title"
// priv.indices.name = name of this index

priv.updateIndices = function (indices, doc) {
var i, j, index, value, label, key, l = priv.indices.length;

// loop all indices to add document
for (i = 0; i < l; i += 1) {
  index = {};
  index.reference = priv.indices[i];
  index.reference_size = index.reference.fields.length;
  index.current = indices[index.reference.name];

  for (j = 0; j < index.reference_size; j += 1) {
    label = index.reference.fields[j];    // like "year"
    value = doc[label];                   // like 1985

    // if document has a label field (e.g. doc.year = 1985)
    if (value !== undefined) {

      // check if the index file already contains an entry for 1985
      index.current_size = priv.getObjectSize(index.current[label]);
      
      if (index.current_size > 0) {
        // check if the document id is already in the index
        // in case the data is updated (e.g. change 1982 to 1985)
        key = priv.searchIndexByValue(
          index.current[label],
          doc._id,
          "key"
        );
        if (!!key) {
          delete index.current[label][key];
        }
      }
      // create a new array if 1985 is not in the index yet
      if (index.current[label][value] === undefined) {
        index.current[label][value] = [];
      }
      // add the document id to an existing entry
      index.current[label][value].push(doc._id);
    }
  }
}
return indices;
};

This works fine, except that fields I want to store as non-strings (integers, numbers or datetime), like the year in the above example end up as strings in my index.

Question:
Is it at all possible to store "non-string" types in a JSON document? If so, can I also store the key of a key/value pair as a "non-string" element.

If not, would I have to add a parameter to my index definitions declaring the type of each key in order to modify the key-string when I run into it or is there a better way to do it?

Thanks!

• javascript
• jquery
• json
• types
• indexing

• 1 according to the json spec, you can have a number anywhere you could have a value. So the key of an object must be a string, but the value can be a number. Also any of the values in an array can be a number. – jbabey Commented Feb 4, 2013 at 13:13

3 Answers 3

Is it at all possible to store "non-string" types in a JSON document?

Yes. The value of a property can be a string, number, boolean, object, array or null (undefined is a notable exception - it's a native JavaScript type but it's not a valid JSON value).

Can I also store the key of a key/value pair as a "non-string" element?

No. The key name must always be a string. However, that doesn't mean you can't parse that string into some other JavaScript type. For example, if you have a string but need a number, you can use the parseInt function, or the unary + operator.

See the JSON grammar for more detail.

no you can't, in JSON keys are strings.

the best you can do is storing string representations of those keys, wether integer or objects(more plicated, you have to build a serialization function).

If you want to use only consecutive integers keys starting from 0, then you can use arrays.

According to the json spec, you can have a number anywhere you could have a value. So the key of an object must be a string, but the value can be a number. Also any of the values in an array can be a number.

The spec is beside the point though; I believe the issue is this line:

index.current[label][value].push(doc._id);

When you read doc._id, that is a string. If you want to store it in the JSON as a number, you need to cast it:

index.current[label][value].push(parseInt(doc._id, 10));

Also note that having just numbers as IDs is not valid HTML.