Unsorted search of combined MongoDB columns - regex

Question

Is it possible to make a search to a virtual column that is composed by two columns?
Let's say I have the following MongoDB collection:
db.collection =
[
{ book : 'The Stand', author : 'Stephen King'},
{ book : 'The Dead Zone', author : 'Stephen King'},
{ book : 'Hamlet', author : 'William Shakespeare'},
{ book : 'The Tragedy of Othello', author : 'William Shakespeare'},
{ book : 'Danse Macabre', author : 'Stephen King'},
]
And I want to make a search that should be made considering both book and author columns at the same time. In particular, I will have a query string with several items separated by spaces, and I would want to return the documents whose joint book+author column contains all the query items regardless of their order.
Example:
Query: "King The"
{ book : 'The Stand', author : 'Stephen King'},
{ book : 'The Dead Zone', author : 'Stephen King'}
Query: "Tragedy Shakespeare"
{ book : 'The Tragedy of Othello', author : 'William Shakespeare'}
Query: "The"
{ book : 'The Stand', author : 'Stephen King'},
{ book : 'The Dead Zone', author : 'Stephen King'},
{ book : 'The Tragedy of Othello', author : 'William Shakespeare'},
Is this kind of search possible in MongoDB? Is there any $regex expression to make it feasible?
Thank you!

Answer 1

Here is an aggregation I think might help...
db.collection.aggregate([
{ $project: { book: 1, author: 1, "book_words": { $split: [ "$book", " " ] }, "author_words": { $split: [ "$author", " " ] } } },
{ $project: { book:1, author: 1, "search_words": { $concatArrays: [ "$book_words", "$author_words" ] } } },
{ $match: { "search_words": { $all: [ "The", "King" ] } } },
{ $project: { "search_words": 0} }
]).pretty()
Explanation:
This aggregation has 4 stages...
$project
$project
$match
$project
The first $project will split the string value in field "book" into an array of words called "book_words", and also split the string value in the field "author" into an array of words called "author_words"
The second $project will concatenate the two new arrays together into a single array called "search_words"
The $match stage filters out records that do not match the search criteria
the final $project stage removes the temporary array field called "search_words"
Resulting documents for this aggregation look like...
{
"_id" : ObjectId("60d6139a9148371ae7d2b343"),
"book" : "The Stand",
"author" : "Stephen King"
}
{
"_id" : ObjectId("60d6139a9148371ae7d2b344"),
"book" : "The Dead Zone",
"author" : "Stephen King"
}
Case insensitive matching
In order to provide case insensitive matching MongoDB must understand what case insensitive means. English case is different from other languages. So for this reason we must add an index with a collation that defines english as the language and a strength of 2 for the collation - meaning case insensitive for english. Once the index is created, we must refer to the collation as a option in the aggregation.
Create Index
db.collection.createIndex( { book: 1, author: 1 }, { collation: { locale: 'en', strength: 2 } } )
This is a compound index on both fields - 'book' and 'author'. Notice collation options for this index...
Aggregation using collation
Now that the index exists with a specific collation, Mongo now can calculate the case insensitive options...
db.collection.aggregate([
{ $project: { book: 1, author: 1, "book_words": { $split: [ "$book", " " ] }, "author_words": { $split: [ "$author", " " ] } } },
{ $project: { book:1, author: 1, "search_words": { $concatArrays: [ "$book_words", "$author_words" ] } } },
{ $match: { "search_words": { $all: [ "the", "king" ] } } },
{ $project: { "search_words": 0} }
],
{ collation: { locale: "en", strength: 2 } }).pretty()
Notice the collation option is applied to the aggregation. Also, the aggregation $match stage is now using all lowercase text.
Here is the output...
{
"_id" : ObjectId("60d6139a9148371ae7d2b343"),
"book" : "The Stand",
"author" : "Stephen King"
}
{
"_id" : ObjectId("60d6139a9148371ae7d2b344"),
"book" : "The Dead Zone",
"author" : "Stephen King"
}
Beware
use of regular expressions with collation options will probably not work as expected, at least from an index strategy point of view. In my example I am not using any regular expressions ($regex), and as such it works as expected. But again, this is for exact matches, not partial matches (a.k.a. range queries) such as "Starts with 'ki*'"
MongoDB Atlas Search
If using MongoDB Atlas the use of Atlas Search solves this problem directly, with the exception of common words such as 'the' are omitted.