The most powerful aspect of bookworm is its ability to filter down to a custom set of fields. This is done by setting a number of constraints on search_limits or compare_limits fields.
As described in 4.2, search_limits and compare_limits both take the same syntax: for this section, I'll just use search_limits but understand that they mean both.
Every bookworm may have different metadata fields: these examples are for a fictitious database of books.
A few keys should be supported in all bookworms, though:
Word searches
"word" (which matches, despite its name, any phrase up to the number of supported grams).unigram or bigram can be used as a synonym for word. This is particularly useful in searching for groups.hasword can limit the search to only books that containing the specified word. (Not currently supported, but part of the specification)In progress: It's possible this should be retitled
$haswordfromhaswordto better match the syntax below.
In progress: There should also be a syntax built in to support proximity searches: these are not possible under MySQL, but would be under Solr. I'd suggest something along the lines of
{"near":["foo",5]}where foo is the word and five is the proximity range.
To limit a query categorically, pass an array consisting of the allowed keys. For example, to limit to women and books published in the United States or Germany in 1890, pass
"search_limits":{
"author_gender":["Female"],
"country":["United States","Germany","East Germany","West Germany"],
"publish_year":[1890]
}
Queries bounded in an array are by default treated as an "or" construction within the array: documents matching any limit will be returned.
Each of the elements, on the other hand, is treated as an AND construction: returned documents must match at least one item for EACH of the keys entered.
If a string or numeric value is passed rather than an array, the API will automatically convert that to a single-element array.
An alternate way to express categorical limits is to use a dictionary rather than array, and pass the special "$eq" key as a value pointing to the limitations. The above query could also be written as:
"search_limits":{
"author_gender":{"$eq":"Female"},
"country":{
"$eq":["United States","Germany","East Germany","West Germany"]
},
"publish_year":{"$eq":1890}
}
This long syntax is borrowed from MongoDB. It is verbose and not useful for most queries: but is provided for completeness because more complicated API calls require other keys (such as "$ne") which take the same form.
To define a search by negative matches, you can pass a hash rather than an array and use the special key "$ne". So to search for books published outside the United States or Canada, you would pass:
{"country":{"$ne":["United States","Canada"]}}
An "$ne" limitation will reject any books published in either of those places.
Numeric values can also be searched with range queries. The syntax for this is to pass a hash and use special keys from the following list:
$gte Greater than or equal to$gt Greater than$lte Less than or equal to$gt Less thanSo to limit to books published between 1900 and 1950, either of the following two queries would work:
{"publish_year":{"$lte":1950,"$gte":1900}
{"publish_year":{"$lt":1951,"$gt":1899}
More complicated boolean statements are possible by using the special $or construction. (This syntax is also borrowed from MongoDB). $or points to an array, any element of which might be true. To search for books that are either published in the US or by US-born authors, for example, you might limit as follows:
{"$or":[
{"country":["USA"]},
{"author_birth_country":["USA"]}
]}
$and is the opposite of $or: so the following query:
"search_limits":{"$and":[{"country":["USA"],"author_birth_country":["USA"]}]}
would return all books published both in the United States and by American-born authors. This and-construction will never be necessary in a base-level search, because all grouped limitations are and-queries by default: the above is exactly equivalent to
"search_limits"{"country":["USA"],"author_birth_country":["USA"]}
Like "$eq" (above), "$and" is included for completeness and so that it can specified in deeply nested queries.
These constructions can be nested arbitrarily to create complicated combinations of "and" and "or" queries. To construct a query, for example, that encompasses authors whose political party is the same as the sitting US president's in the last few decades, you might type the following monstrosity:
"search_limits":{
"$or":[
{
"year":{
"$or":[
{"$gte":1980,"$lte":1992},
{"$gte":2001,"$lte":2008}
]
},
"author_party":"Republican"
},
{
"year":{
"$or":[
{"$gte":1993,"$lte":2000},
{"$gte":2009}
]
},
"author_party":"Democrat"
}
]
}
You can also limit by regular expressions rather than complete matches by using the "$grep" key.
MySQL-supported regular expressions can be used. To match any of several spellings of the publisher "Little, Brown" you might search for:
"search_limits":{"publisher":{"$grep":"Little,? Brown ((and|&) ?[Cc]o\.?)?"}
Yuck, huh?
Question: Should this be retitled
$re?
Should it be possible to limit not just by results, but by counttypes? For instance, to only return results where the specified groups have a wordcount over 10, you could search:
"search_limits":{"WordCount":{"$gte":10}}