v7.6.1

This is a bugfix release that addresses a backwards incompatibility in the TypeScript type signatures in v7.6.0 and does not include any changes relevant to non-TypeScript users.

Fixed

• Changed all uses of Record<string, unknown> to Record<string, any> (#750)

  This should allow using more specific types without having to implement
  index signatures.

v7.6.0

Added

• Added collection.documents for fetching multiple documents

• Added support for fillBlockCache query option

• Added support for passing Graph objects in AQL queries (#740)

  This also adds the isArangoGraph helper function for type checking.

• Added User Management API (#664)

  This implements the endpoints of the
  HTTP Interface for User Management

Fixed

• Added missing hex flag to StopwordsAnalyzer type (#732)

• Added missing details flag to collection.figures (#728)

• Added missing inBackground flag to index options (#734)

v7.5.0

Added

• Added support for new ArangoDB 3.8 Analyzer types

  This adds the PipelineAnalyzer, AqlAnalyzer, GeoJsonAnalyzer,
  GeoPointAnalyzer and StopwordsAnalyzer types in TypeScript, as well as
  the Analyzer-specific properties types.

• Added support for new ArangoDB 3.8 estimates option for indexes

  This affects the PersistentIndex, HashIndex and SkiplistIndex types
  in TypeScript.

v7.5.0-preview-arangodb.3.8

This is a preview release which is not intended for use in production and has been published under the npm next tag.

To install the latest preview release, run npm install arangojs@next or yarn add arangojs@next.

Added

• Added support for new ArangoDB 3.8 Analyzer types

  This adds the PipelineAnalyzer, AqlAnalyzer, GeoJsonAnalyzer,
  GeoPointAnalyzer and StopwordsAnalyzer types in TypeScript, as well as
  the Analyzer-specific properties types.

• Added support for new ArangoDB 3.8 estimates option for indexes

  This affects the PersistentIndex, HashIndex and SkiplistIndex types
  in TypeScript.

v7.4.0

Added

• Implemented toJSON methods for ArangoError and HttpError (#632)

  This prevents an error where JSON.stringify would reliably throw if passed
  an instance of either of these error types generated by arangojs. Note that
  you may still want to implement your own JSON representation logic as system
  errors (e.g. ECONNREFUSED) are not wrapped by arangojs and thrown as-is.

Fixed

• Stack traces are now improved for most errors when using precaptureStackTraces (#722)

  Previously this option would only affect network errors, making it far less
  useful than intended. Now parsing errors, ArangoError instances and HTTP
  errors also receive improved error stack traces when this option is enabled.

• Improved performance for precaptureStackTraces when no errors occur

  The generated stack is now only accessed on demand, allowing the runtime to
  delay generation of the stack trace string. Previously the stack would always
  be accessed prior to the request being sent, causing a noticeable delay even
  when no error occurs.

• Fixed document selector validation in collection.edges and its variants (#704)

  These methods previously only permitted start vertices that are documents
  within the edge collection itself. This behavior has now been corrected to
  permit start vertices outside the collection, as expected.

v7.3.0

Changed

• Changed the default for agentOptions.scheduling to "lifo"

  This is already the default in Node v15.6 but can reduce latency caused by
  sockets expiring, especially with larger connection pools and infrequent
  requests.

• Removed keepAlive-specific throughput optimization

  Previously arangojs would allow agentOptions.maxSockets * 2 concurrent
  requests, to optimize socket reuse by avoiding idle time. This behavior
  could trigger deadlocks when attempting to perform multiple transactions
  in parallel and only marginally improved throughput in some high-load
  scenarios. The connection pool size now always reflects the value set in
  agentOptions.maxSockets regardless of whether keepAlive is enabled.

• Changed agentOptions.maxSockets default value when using ROUND_ROBIN

  As the connection pool is shared across all server connections when using
  ROUND_ROBIN load balancing, the default value of 3 is too limiting for
  most scenarios involving multiple coordinators. When passing multiple URLs
  via the url option and specifying ROUND_ROBIN load balancing, arangojs
  will now default this value to url.length * 3 instead.

v7.2.0

Added

• Added db.waitForPropagation method

  This method helps with setting up databases in a cluster scenario by waiting
  for a request to succeed on every known coordinator.

v7.1.1

This is a maintenance release and contains no bugfixes or features.

v7.1.0

Changed

• Killing a cursor now also drains it locally

Fixed

• Fixed a potential memory leak in cursor batch handling

v7.0.2

Fixed

• Fixed incorrect HTTP method call in patch method (#687)

• Fixed empty query results containing [undefined] (#683)

• Fixed updateByExample and replaceByExample new value parameter name

  Note that these methods are still deprecated. Previously the newValue
  parameter was incorrectly called newData, which prevented the methods from
  working at all.