Skip to main content
Search tokens are a powerful mechanism in Orionjs that enable efficient and flexible text search capabilities in MongoDB without the overhead of full-text search. They work by preprocessing text fields into normalized tokens that can be indexed and queried efficiently.

​
Why Use Search Tokens?

  • Simplicity: No need to create complex regex queries or text indexes
  • Performance: Significantly faster than regex or text queries
  • Flexibility: Combine text search with category filtering
  • Normalized Search: Case-insensitive and accent-insensitive matching
  • Prefix Matching: Find results that start with search terms

​
Implementation

​
1. Add Search Tokens Field to Your Schema

First, add a searchTokens field to your schema: 
import {schemaWithName, InferSchemaType} from '@orion-js/schema'
import {typedId} from '@orion-js/mongodb'

export const UserSchema = schemaWithName('UserSchema', {
  _id: {type: typedId('usr')},
  firstName: {type: String},
  lastName: {type: String},
  email: {type: String},
  role: {type: String},
  searchTokens: {type: [String]}
})

export type UserType = InferSchemaType<typeof UserSchema>

​
2. Create an Index on Search Tokens

Add an index on the searchTokens field in your repository: 
import {Repository} from '@orion-js/services'
import {createCollection} from '@orion-js/mongodb'
import {UserSchema, UserType} from '../schemas/UserSchema'

@Repository()
export class UserRepository {
  users = createCollection({
    name: 'users',
    schema: UserSchema,
    indexes: [
      {
        keys: {
          searchTokens: 1,
        }
      }
    ]
  })
}

​
3. Implement a Method to Generate Search Tokens

Add a method to generate search tokens from relevant fields: 
private getSearchTokens(doc: UserType) {
  // Text fields to tokenize
  const texts = [
    doc.firstName,
    doc.lastName,
    doc.email,
    shortenMongoId(doc._id) // Include ID in searchable fields
  ]
  
  // Category fields (for filtering)
  return getSearchTokens(texts, {
    role: doc.role
  })
}

​
4. Update Search Tokens When Creating or Updating Documents

async createUser(doc: Omit<UserType, '_id'>) {
  const docWithTokens = {
    ...doc,
    searchTokens: this.getSearchTokens({...doc, _id: `usr-temp`} as UserType)
  }
  return await this.users.insertAndFind(docWithTokens)
}

async updateUser(userId: string, doc: Partial<UserType>) {
  const result = await this.users.updateAndFind(userId, {$set: doc})
  this.ensureSearchTokensAreUpdated(result)
  return result
}

private ensureSearchTokensAreUpdated(user: UserType) {
  const searchTokens = this.getSearchTokens(user)

  if (!isEqual(user.searchTokens, searchTokens)) {
    this.users
      .updateOne(user._id, {
        $set: {searchTokens},
      })
      .catch(error => {
        logger.error('Error updating search tokens', {error, userId: user._id, searchTokens})
      })
  }
}

​
5. Query Using Search Tokens

private async getUsersListQuery(params: UsersListQueryParams) {
  const queries: MongoFilter<UserType>[] = []

  const searchTokensParams: Record<string, string> = {}

  if (params.role) {
    searchTokensParams.role = params.role
  }

  const searchTokens = getSearchQueryForTokens({
    filter: params.filter,
    ...searchTokensParams,
  })

  if (searchTokens.$all.length) {
    queries.push({searchTokens})
  }

  return {
    $and: queries,
  }
}

async getUsersListCursor(params: UsersListQueryParams) {
  const query = await this.getUsersListQuery(params)
  return this.users.find(query)
}

​
How Search Tokens Work

  1. Text Tokenization: Text fields are split into tokens, converted to lowercase, and normalized
  2. Prefix Generation: Additional tokens are created for prefixes to enable prefix searching
  3. Category Markers: Category fields are converted to tokens with prefixes to enable category filtering
  4. Query Building: The getSearchQueryForTokens function converts search terms into MongoDB queries

​
Best Practices

  • Include Important Text Fields: Add all searchable text fields to the tokens
  • Short MongoDB IDs: Use shortenMongoId to include readable portions of IDs
  • Category Fields: Include fields used for filtering in the second argument of getSearchTokens
  • Ensure Tokens are Updated: Always update search tokens when document fields change
  • Checking Token Equality: Use a deep comparison like isEqual to avoid unnecessary updates
  • Error Handling: Implement proper error handling for token updates
  • Background Updates: Update tokens in the background to avoid blocking user operations

​
Complete Example

import {Repository} from '@orion-js/services'
import {createCollection, MongoFilter, typedId} from '@orion-js/mongodb'
import {schemaWithName, InferSchemaType} from '@orion-js/schema'
import {UsersListQueryParams} from './ListQueryParams'
import {isEqual} from 'lodash'
import {logger} from '@orion-js/logger'
import {getSearchTokens, getSearchQueryForTokens, shortenMongoId} from '@orion-js/helpers'

const UserSchema = schemaWithName('UserSchema', {
  _id: {type: typedId('usr')},
  firstName: {type: String},
  lastName: {type: String},
  email: {type: String},
  role: {type: String},
  status: {type: String},
  searchTokens: {type: [String]}
})

type UserType = InferSchemaType<typeof UserSchema>

@Repository()
export class UsersRepo {
  users = createCollection({
    name: 'users',
    schema: UserSchema,
    indexes: [
      {
        keys: {
          searchTokens: 1,
        }
      }
    ]
  })

  private getSearchTokens(doc: UserType) {
    const texts = [
      doc.firstName,
      doc.lastName,
      doc.email,
      shortenMongoId(doc._id)
    ]
    return getSearchTokens(texts, {
      role: doc.role,
      status: doc.status
    })
  }

  async createUser(doc: Omit<UserType, '_id'>) {
    const docWithTokens = {
      ...doc,
      searchTokens: this.getSearchTokens({...doc, _id: `usr-temp`} as UserType)
    }
    return await this.users.insertAndFind(docWithTokens)
  }

  async updateUser(userId: string, doc: Partial<UserType>) {
    const result = await this.users.updateAndFind(userId, {$set: doc})
    this.ensureSearchTokensAreUpdated(result)
    return result
  }

  private ensureSearchTokensAreUpdated(user: UserType) {
    const searchTokens = this.getSearchTokens(user)

    if (!isEqual(user.searchTokens, searchTokens)) {
      this.users
        .updateOne(user._id, {
          $set: {searchTokens},
        })
        .catch(error => {
          logger.error('Error updating search tokens', {error, userId: user._id, searchTokens})
        })
    }
  }

  private async getUsersListQuery(params: UsersListQueryParams) {
    const queries: MongoFilter<UserType>[] = []

    const searchTokensParams: Record<string, string> = {}

    if (params.role) {
      searchTokensParams.role = params.role
    }

    if (params.status) {
      searchTokensParams.status = params.status
    }

    const searchTokens = getSearchQueryForTokens({
      filter: params.filter,
      ...searchTokensParams,
    })

    if (searchTokens.$all.length) {
      queries.push({searchTokens})
    }

    return {
      $and: queries.length ? queries : [{}],
    }
  }

  async getUsersListCursor(params: UsersListQueryParams) {
    const query = await this.getUsersListQuery(params)
    return this.users.find(query)
  }
}

​
Performance Considerations

  • Keep the number of tokens reasonable (< 100 per document)
  • Consider sharding for very large collections
  • For extremely complex search needs, consider using a dedicated search engine
⌘I