ArangoSearch Examples

Assume the following collections were initially defined in a database usingthe following arangosh commands:

  1. c0 = db._create("ExampleCollection0");
  2. c1 = db._create("ExampleCollection1");
  4. c0.save({ i: 0, name: "full", text: "是一个 多模 型数 据库" });
  5. c0.save({ i: 1, name: "half", text: "是一个 多模" });
  6. c0.save({ i: 2, name: "other half", text: "型数 据库" });
  7. c0.save({ i: 3, name: "quarter", text: "是一" });
  9. c1.save({ a: "foo", b: "bar", i: 4 });
  10. c1.save({ a: "foo", b: "baz", i: 5 });
  11. c1.save({ a: "bar", b: "foo", i: 6 });
  12. c1.save({ a: "baz", b: "foo", i: 7 });

A View with default parameters can be created by calling db._createView()(see JavaScript Interface for Views):

  1. v0 = db._createView("ExampleView", "arangosearch", {});

Then link the collections to the View, define which properties shall be indexedand with which Analyzers the values are supposed to be processed:

  1. v0 = db._view("ExampleView");
  2. v0.properties({
  3.     links: {
  4.       /* collection Link 0 with additional custom configuration: */
  5.       'ExampleCollection0':
  6.       {
  7.         /* examine fields of all linked collections,
  8.            using default configuration: */
  9.         includeAllFields: true,
  10.         fields:
  11.         {
  12.           /* a field to apply custom configuration
  13.              that will index English text: */
  14.           name:
  15.           {
  16.             analyzers: ["text_en"]
  17.           },
  18.           /* another field to apply custom configuration
  19.              that will index Chinese text: */
  20.           text:
  21.           {
  22.             analyzers: ["text_zh"]
  23.           }
  24.         }
  25.       },
  26.       /* collection Link 1 with custom configuration: */
  27.       'ExampleCollection1':
  28.       {
  29.         /* examine all fields using default configuration: */
  30.         includeAllFields: true,
  31.         fields:
  32.         {
  33.           a:
  34.           {
  35.             /* a field to apply custom configuration
  36.                 that will index English text: */
  37.             analyzers: ["text_en"]
  38.           }
  39.         }
  40.       }
  41.     }
  42.   }
  43. );

You have to wait a few seconds for the View to update its index.Then you can query it:

  1. db._query(`FOR doc IN ExampleView
  2.   SEARCH PHRASE(doc.text, "型数 据库", "text_zh") OR
  3.   ANALYZER(STARTS_WITH(doc.b, "ba"), "text_en")
  4.   SORT TFIDF(doc) DESC
  5.   RETURN doc`);

The result will include all documents from both linked collections that havethe Chinese phrase 多模 型数 at any position in the text attribute valueor the property b starting with ba in English. Additionally,descendant sorting using the TF-IDF scoring algorithm will be applied.

  1. [
  2.   {
  3.     "_key" : "120",
  4.     "_id" : "ExampleCollection0/120",
  5.     "_rev" : "_XPoMzCi--_",
  6.     "i" : 0,
  7.     "name" : "full",
  8.     "text" : "是一个 多模 型数 据库"
  9.   },
  10.   {
  11.     "_key" : "124",
  12.     "_id" : "ExampleCollection0/124",
  13.     "_rev" : "_XPoMzCq--_",
  14.     "i" : 2,
  15.     "name" : "other half",
  16.     "text" : "型数 据库"
  17.   },
  18.   {
  19.     "_key" : "128",
  20.     "_id" : "ExampleCollection1/128",
  21.     "_rev" : "_XPoMzCu--_",
  22.     "a" : "foo",
  23.     "b" : "bar",
  24.     "c" : 0
  25.   },
  26.   {
  27.     "_key" : "130",
  28.     "_id" : "ExampleCollection1/130",
  29.     "_rev" : "_XPoMzCy--_",
  30.     "a" : "foo",
  31.     "b" : "baz",
  32.     "c" : 1
  33.   }
  34. ]

Use cases

The data contained in our View looks like that:

  1. { "id": 1, "body": "ThisIsAVeryLongWord" }
  2. { "id": 2, "body": "ThisIsNotSoLong" }
  3. { "id": 3, "body": "ThisIsShorter" }
  4. { "id": 4, "body": "ThisIs" }
  5. { "id": 5, "body": "ButNotThis" }

We now want to search for documents where the attribute body starts with"ThisIs". A simple AQL query executing this prefix search:

  1. FOR doc IN viewName
  2.   SEARCH STARTS_WITH(doc.body, 'ThisIs')
  3.   RETURN doc

It will find the documents with the ids 1, 2, 3, 4, but not 5.