BACKWARD-INCOMPATIBLE: Added multiple index support to Haystack, which enables you to talk to more than one search engine in the same codebase. Thanks to:

* Kent Gormat for funding the development of this feature.
* alex, freakboy3742 & all the others who contributed to Django's multidb feature, on which much of this was based.
* acdha for inspiration & feedback.
* dcramer for inspiration & feedback.
* mcroydon for patch review & docs feedback.

This commit starts the development efforts for Haystack v2.

Author: toastdriven

TODO

@@ -45,9 +45,9 @@
 # built documents.
 #
 # The short X.Y version.
-version = '1.2'
+version = '2.0'
 # The full version, including alpha/beta/rc tags.
-release = '1.2.0'
+release = '2.0.0-alpha'
 
 # The language for content autogenerated by Sphinx. Refer to documentation
 # for a list of supported languages.

docs/conf.py

@@ -77,14 +77,14 @@ small sites or development.
 
 A way to solve this is to subclass ``SearchIndex`` instead::
 
-    from haystack.indexes import *
+    from haystack import indexes
     
     # Change from:
     # 
-    #   class MySearchIndex(RealTimeSearchIndex):
+    #   class MySearchIndex(indexes.RealTimeSearchIndex):
     # 
     # to:
-    class MySearchIndex(SearchIndex):
+    class MySearchIndex(indexes.SearchIndex):
         ...
 
 The final step is to set up a cron job that runs
@@ -97,52 +97,16 @@ traffic. If this is not acceptable, you should investigate either the Solr or
 Xapian backends.
 
 
-"Import errors on start-up mentioning 'handle_registrations'"
-=============================================================
-
-When initializing, Haystack attempts to import and register all of the
-``SearchIndex`` classes you've setup. As a by-product of this, especially in
-conjunction with third-party apps that attempt to do similar types of imports,
-it's possible (though rare) to get a traceback very early in the start-up
-process, usually mentioning ``handle_registrations``.
-
-There are typically three possible causes for this error:
-
-    * A syntax/import error in a file included by the ``search_indexes.py`` file
-    * A circular import
-    * Another app causing models to load early
-
-The first two causes can be debugged by dropping an
-``import pdb; pdb.set_trace()`` at the top of the ``search_indexes.py`` where
-the error is occurring and stepping through to see the real error.
-
-If neither of those is the case, Haystack provides an advanced setting
-(``HAYSTACK_ENABLE_REGISTRATIONS`` - :doc:`settings`) to disable this importing
-behavior and allow your applications to function.
-
-As a note of caution, setting ``HAYSTACK_ENABLE_REGISTRATIONS = False`` in your
-settings causes Haystack to be left in an uninitialized state. This means that
-none of your ``SearchIndex`` classes will be registered and all attempts to use
-``SearchQuerySet`` will yield no results. To continue using Haystack, you'll
-need to manually import your ``search_indexes.py`` files, either in your
-``models.py`` or ``views.py`` files (or something similar). Additionally, any
-use at the console/management commands may also require similar imports.
-
-Finally, should this occur to you, it would be appreciated if you could report
-the issue and the app(s) you're using that are causing the issue in conjunction
-with Haystack on either the mailing list or on the GitHub issue tracker.
-
-
 "Failed to add documents to Solr: [Reason: None]"
 =================================================
 
 This is a Solr-specific traceback. It generally occurs when there is an error
-with your ``HAYSTACK_SOLR_URL``. Since Solr acts as a webservice, you should
+with your ``HAYSTACK_CONNECTIONS[<alias>]['URL']``. Since Solr acts as a webservice, you should
 test the URL in your web browser. If you receive an error, you may need to
 change your URL.
 
 This can also be caused when using old versions of pysolr (2.0.9 and before),
-using httplib2 and including a trailing slash in your ``HAYSTACK_SOLR_URL``.
+using httplib2 and including a trailing slash in your ``HAYSTACK_CONNECTIONS[<alias>]['URL']``.
 Please upgrade your version of pysolr (2.0.13+).
 
 

docs/debugging.rst

@@ -41,6 +41,8 @@ you up and running:
    other_apps
    installing_search_engines
    debugging
+   
+   migration_from_1_to_2
 
 
 Advanced Uses
@@ -57,7 +59,7 @@ you may want to include in your application.
    faceting
    autocomplete
    boost
-   do_not_try_this_at_home
+   multiple_index
 
 
 Reference
@@ -73,7 +75,6 @@ looking for API documentation and advanced usage as detailed in:
    searchindex_api
    searchfield_api
    searchresult_api
-   searchsite_api
    searchquery_api
    searchbackend_api
 

docs/do_not_try_this_at_home.rst

@@ -20,14 +20,17 @@ following arguments::
         uncerimoniously wiped out.
     ``--verbosity``:
         Accepted but ignored.
+    ``--using``:
+        If provided, determines which connection should be used. Default is
+        ``default``.
 
 By default, this is an **INTERACTIVE** command and assumes that you do **NOT**
 wish to delete the entire index.
 
 .. warning::
 
   Depending on the backend you're using, this may simply delete the entire
-  directory, so be sure your ``HAYSTACK_<ENGINE>_PATH`` setting is correctly
+  directory, so be sure your ``HAYSTACK_CONNECTIONS[<alias>]['PATH']`` setting is correctly
   pointed at just the index directory.
 
 
@@ -58,6 +61,9 @@ arguments::
             and how many records.
           * ``2`` = Full output, including everything from ``1`` plus output
             on each batch that is indexed, which is useful when debugging.
+    ``--using``:
+        If provided, determines which connection should be used. Default is
+        ``default``.
 
 .. note::
 
@@ -96,6 +102,9 @@ of the arguments of the following arguments::
             and how many records.
           * ``2`` = Full output, including everything from ``1`` plus output
             on each batch that is indexed, which is useful when debugging.
+    ``--using``:
+        If provided, determines which connection should be used. Default is
+        ``default``.
 
 For when you really, really want a completely rebuilt index.
 
@@ -104,12 +113,19 @@ For when you really, really want a completely rebuilt index.
 =====================
 
 Once all of your ``SearchIndex`` classes are in place, this command can be used
-to generate the XML schema Solr needs to handle the search data. It accepts no
-arguments.
+to generate the XML schema Solr needs to handle the search data. It accepts the
+following arguments::
+
+    ``--filename``:
+        If provided, directs output to a file instead of stdout.
+    ``--using``:
+        If provided, determines which connection should be used. Default is
+        ``default``.
 
 .. warning:
 
-    This command does NOT update the ``schema.xml`` file for you. You have to
+    This command does NOT update the ``schema.xml`` file for you. You either
+    have to specify a ``filename`` flag or have to
     copy-paste (or redirect) the output to the correct file. Haystack has no
     way of knowing where your Solr is setup (or if it's even on the same
     machine), hence the manual step.