wildcard.hps

CastleCMS and Plone integration with OpenSearch

This product was forked from collective.elasticsearch in order to provide integration with OpenSearch instead of ElasticSearch. OpenSearch itself is a fork of ElasticSearch and compatible with, at least, the ES 7.10.x series of releases (at least at opensearch-py 1.1.0). Compatibility may diverge in the future, and while the collective.elasticsearch package will likely try to maintain compatibility with ElasticSearch, wildcard.hps is intended to maintain compatibility with OpenSearch.

Quickstart

First, start up an instance (for official guides, see the opensearch project documentation)

$ docker run -p 9200:9200 -p 9600:9600 -e "discovery.type=single-node" opensearchproject/opensearch:latest
$ curl -XGET https://localhost:9200 -u 'admin:admin' -k

Second, setup Plone/CastleCMS:

1. add wildcard.hps to the eggs section of your buildout
2. run buildout
3. restart your instance, using relevant Environment Variables to connect to your opensearch instance
4. install the 'Wildcard HPS' product
5. under the 'Wildcard HPS' control panel, click 'Convert Catalog' then 'Rebuild Catalog'

Configuration Settings are passed as environment variables. See the "Configuration" section below for more details.

Overview

This package aims to index all fields the portal_catalog indexes and allows you to delete the Title, Description and SearchableText indexes which can provide significant improvement to performance and RAM usage.

OpenSearch queries are ONLY used when Title, Description and SearchableText text are in the query. Otherwise, Plone's default catalog will be used. This is because Plone's default catalog is faster on normal queries than using OpenSearch.

Configuration

Configuration for OpenSearch connections, and custom index naming, is done through Environment Variables. This allows per-instance customization without the need to modify site data, and allows for many deployments to use the same cluster(s) without needing to do per-site customized index names.

Available Environrment Variable Options:

• HPS_ZOPE_CONF_PATH
  • path to a zope.conf to get a Zope app instance
  • NOTE: this is only needed for the reindex_hps script that gets installed. See wildcard/hps/scripts/reindex.py.
• HPS_OVERRIDE_LOGGING
  • if present, will tell the reindex_hps script to override the root logging configuration, and print logging to console at INFO level.
• HPS_FORCE_ENABLE
  • default: no
  • accepted values (all other values are equivalent to False): Yes, True, 1, On
  • will force the "enabled" lookup to be True
• HPS_INSTANCE_INDEX_PREFIX
  • default: None
  • a string value prepended to index names used by the Plone instances this addon is installed into
• HPS_INCLUDE_TRASHED_BY_DEFAULT
  • default: no
  • accepted values (all other values are equivalent to False): Yes, True, 1, On
  • will default searchResults to include trashed entries (which are not included by default)
• HPS_FOCE_EXTERNAL_INDEXES
  • default: None
  • a list of object properties that will be included in the externally index object (IE the indexed object in opensearch)
• OPENSEARCH_HOSTS
  • default: https://admin:admin@localhost:9200
  • a list of RFC-1738 formated urls. multiple urls can be specified by putting a space between urls.
  • NOTE: for now, the opensearch-py (1.1.0) does not respect the HTTP auth info that is formatted as part of the URL, instead use OPENSEARCH_HTTP_USERNAME and OPENSEARCH_HTTP_PASSWORD to pass the same HTTP auth to each request to any node listed as a host.
• OPENSEARCH_HTTP_USERNAME
  • default: None
  • a username to use in all connections to any node in the OPENSEARCH_HOSTS list
• OPENSEARCH_HTTP_PASSWORD
  • default: None
  • a password to use in all connections to any node in the OPENSEARCH_HOSTS list
• OPENSEARCH_TIMEOUT
  • default connection timeout
• OPENSEARCH_RETRY_ON_TIMEOUT
  • default: Off
  • accepted values (all other values are equivalent to False): Yes, True, 1, On
  • retry connection to different node when connection fails
• OPENSEARCH_SNIFF_ON_START
  • default: False
  • accepted values (all other values are equivalent to False): Yes, True, 1, On
  • refresh nodes before doing anything
• OPENSEARCH_SNIFF_ON_CONNECTION_FAIL
  • default: False
  • accepted values (all other values are equivalent to False): Yes, True, 1, On
  • refresh nodes after a node fails to respond
• OPENSEARCH_SNIFFER_TIMEOUT
  • default: None
  • refresh node list on this time (in seconds) interval
• OPENSEARCH_SNIFF_TIMEOUT
  • default: 0.1
  • timeout of sniff request
• OPENSEARCH_USE_SSL
  • default: False
  • accepted values (all other values are equivalent to False): Yes, True, 1, On
  • connections to OpenSearch will use SSL
• OPENSEARCH_VERIFY_CERTS
  • default: True
  • accepted values (all other values are equivalent to False): Yes, True, 1, On
  • verify SSL certificates when using SSL connections to OpenSearch
• OPENSEARCH_SSL_SHOW_WARN
  • default: True
  • accepted values (all other values are equivalent to False): Yes, True, 1, On
  • when verifying SSL certificates is disabled, then a warning will be shown by default
• OPENSEARCH_CERTS_PATH
  • default: None
  • a path to a directory containing CA Certificates used in SSL verification
• OPENSEARCH_CLIENT_CERT_PATH
  • default: None
  • a path to a PEM formated SSL client certificate for SSL client auth
• OPENSEARCH_CLIENT_CERT_KEY --
  • default: None
  • a path to a PEM formated SSL client key for SSL client auth

Compatibility

Only tested with Plone 5 with Dexterity types.

Only compatible with versions of OpenSearch (and ElasticSearch) compatible with the opensearch-py library.

For ElasticSearch integration, see collective.elasticsearch.

State

Support for all index column types is done EXCEPT for the DateRecurringIndex index column type. If you are doing a full text search along with a query that contains a DateRecurringIndex column, it will not work.

Celery support

This package comes with Celery support where all indexing operations will be pushed into celery to be run asynchronously.

Please see instructions for collective.celery to see how this works.

Running tests

First, start an instance of OpenSearch.

Second,

$ virtualenv ./env
$ ./env/bin/pip install -r requirements.txt
$ ./env/bin/buildout -c buildout.cfg
$ ./bin/test

Changelog

1.4.3 (2023-10-11)

• handle unicode for index data derived from IAdditionalIndexDataProvider adapters

1.4.2 (2023-05-15)

• abstract unicode handling code for hook when getting index data, and handle tuples, lists, and dict values

1.4.1 (2023-05-11)

• handle unicode error and fix bug in hook when getting index data

1.4.0 (2022-11-04)

• allow a custom prefix to be defined for fetching connection settings from the environment (default to the previous hard-coded 'OPENSEARCH_' value)

1.3.0 (2022-08-17)

• add HPS_FORCE_EXTERNAL_INDEXES
• update default set returned when external indexes setting is not configured yet

1.2.1 (2022-06-23)

• fix some view name's in the control panel templates

1.2.0 (2022-05-25)

• add HPS_INCLUDE_TRASHED_BY_DEFAULT env for disabling a filter on searchResults from WildcardHPSCatalog (see readme entry for HPS_INCLUDE_TRASHED_BY_DEFAULT)

1.1.1 (2022-05-12)

• add property on wildcard.hps.opensearch.WildcardHPSCatalog for the instance prefix

1.1.0 (2022-05-12)

• initial fork from: https://github.com/collective/collective.elasticsearch/commit/d21bf7b9311a9fc923283eeff11c42f4145180b4 this fork aims to primarily maintain compatibility with the OpenSearch project, which itself has forked from ElasticSearch 7.10.