NAME

Catalyst::Plugin::AutoCRUD - Instant AJAX web front-end for DBIx::Class

VERSION

version 1.110470

SYNOPSIS

If you already have a Catalyst app with DBIx::Class models configured:

use Catalyst qw(AutoCRUD); # <-- add the plugin name here in MyApp.pm

Now load your app in a web browser, but add /autocrud to the URL path.

Alternatively, to connect to an external database if you have the DBIX::Class schema available, use the ConfigLoader plugin with the following config:

<Model::AutoCRUD::DBIC>
    schema_class   My::Database::Schema
    connect_info   dbi:Pg:dbname=mydbname;host=mydbhost.example.com;
    connect_info   username
    connect_info   password
    <connect_info>
        AutoCommit   1
    </connect_info>
</Model::AutoCRUD::DBIC>

If you don't have the DBIx::Class schema available, just omit the schema_class option (and have DBIx::Class::Schema::Loader installed).

DESCRIPTION

This module contains an application which will automatically construct a web interface for a database on the fly. The web interface supports Create, Retrieve, Update, Delete and Search operations.

The interface is not written to static files on your system, and uses AJAX to act upon the database without reloading your web page (much like other Web 2.0 applications, for example Google Mail).

Almost all the information required by the plugin is retrieved from the DBIx::Class ORM frontend to your database, which it is expected that you have already set up (although see "USAGE", below). This means that any change in database schema ought to be reflected immediately in the web interface after a page refresh.

PURPOSE

You have a database, and wish to have a basic web interface supporting Create, Retrieve, Update, Delete and Search, with little effort. This module is able to create such interfaces on the fly. They are a bit whizzy and all Web 2.0-ish.

USAGE

Read Me First

If you created your DBIx::Class Schema some time ago, perhaps using an older version of DBIx::Class::Schema::Loader, then it might well be lacking some configuration which is required to get the best results from this plugin.

Common omissions in column configurations include is_foreign_key, join_type, is_nullable, and is_auto_increment. Of course it's also good practice to have your DBIx::Class Schema closely reflect the database schema anyway.

To automatically bring things up to date, download the latest version of DBIx::Class::Schema::Loader from CPAN (which may be 0.05 or a pre-release), and use the output from that. If you don't yet have a Schema, continue reading and Scenario 2, below, will cover the steps required.

Scenario 1: Plugin to an existing Catalyst App

This mode is for when you have written your Catalyst application, but the Views are catering for the users and as an admin you'd like a more direct, secondary web interface to the database.

package AutoCRUDUser;
use Catalyst qw(AutoCRUD);

__PACKAGE__->setup;
1;

Adding Catalyst::Plugin::AutoCRUD as a plugin to your Catalyst application, as above, causes it to scan your existing Models. If any of them are built using Catalyst::Model::DBIC::Schema, they are automatically loaded.

This mode of operation works even if you have more than one database. You will be offered a Home screen to select the database, and then another menu to select the table within that.

Remember that the pages available from this plugin will be located under the /autocrud path of your application. Use the basepath option if you want to override this.

Scenario 2: Frontend for an existing `DBIx::Class::Schema` based class

In this mode, Catalyst::Plugin::AutoCRUD is running standalone, in a sense as the Catalyst application itself. Your main application file looks almost the same as in Scenario 1, except you'll need the ConfigLoader plugin:

package AutoCRUDUser;
use Catalyst qw(ConfigLoader AutoCRUD);

__PACKAGE__->setup;
1;

For the configuration, you need to tell AutoCRUD which package contains the DBIx::Class schema, and also provide database connection parameters.

<Model::AutoCRUD::DBIC>
    schema_class   My::Database::Schema
    connect_info   dbi:Pg:dbname=mydbname;host=mydbhost.example.com;
    connect_info   username
    connect_info   password
    <connect_info>
        AutoCommit   1
    </connect_info>
</Model::AutoCRUD::DBIC>

The Model::AutoCRUD::DBIC section must look (and be named) exactly like that above, except you should of course change the schema_class value and the values within connect_info.

Remember that the pages available from this plugin will be located under the /autocrud path if your application. Use the basepath option if you want to override this.

`DBIx::Class` setup

You will of course need the DBIx::Class schema to be created and installed on your system. The recommended way to do this quickly is to use the excellent DBIx::Class::Schema::Loader module which connects to your database and writes DBIx::Class Perl modules for it.

Pick a suitable namespace for your schema, which is not related to this application. For example DBIC::Database::Foo::Schema for the Foo database (in the configuration example above we used My::Database::Schema). Then use the following command-line incantation:

perl -MDBIx::Class::Schema::Loader=make_schema_at,dump_to_dir:. -e \
    'make_schema_at("DBIC::Database::Foo::Schema", { debug => 1 }, \
    ["dbi:Pg:dbname=foodb;host=mydbhost.example.com","user","pass" ])'

This will create a directory (such as DBIC) which you need to move into your Perl Include path (one of the paths shown at the end of perl -V).

Scenario 3: Lazy loading a `DBIx::Class` schema

If you're in such a hurry that you can't create the DBIx::Class schema, as shown in the previous section, then Catalyst::Plugin::AutoCRUD is able to do this on the fly, but it will slow the application's startup just a little.

The application file and configuration are very similar to those in Scenario two, above, except that you omit the schema_class configuration option because you want AutoCRUD to generate that on the fly (rather than reading an existing one from disk).

package AutoCRUDUser;
use Catalyst qw(ConfigLoader AutoCRUD);

__PACKAGE__->setup;
1;

<Model::AutoCRUD::DBIC>
    connect_info   dbi:Pg:dbname=mydbname;host=mydbhost.example.com;
    connect_info   username
    connect_info   password
    <connect_info>
        AutoCommit   1
    </connect_info>
</Model::AutoCRUD::DBIC>

When AutoCRUD loads it will connect to the database and use the DBIx::Class::Schema::Loader module to reverse engineer its schema. To work properly you'll need the very latest version of that module (at least 0.05, or the most recent development release from CPAN).

The other drawback to this scenario (other than the slower operation) is that you have no ability to customize how foreign, related records are shown. A related record will simply be represented as something approximating the name of the foreign table, the names of the primary keys, and associated values (e.g. id(5)).

TIPS AND TRICKS

Update your `DBIx::Class` Result Classes

To automatically bring things up to date, download the latest version of DBIx::Class::Schema::Loader from CPAN (which may be 0.05 or a pre-release), and use the output from that.

More detail is given in the "TROUBLESHOOTING" section, below.

When the web interface wants to display a column which references another table, you can make things look much better by adding a custom render method to your DBIx::Class Result Classes (i.e. the class files for each table).

First, the plugin will look for a method called display_name and use that. Here is an example which could be added to your Result Class files below the line which reads DO NOT MODIFY THIS OR ANYTHING ABOVE, and in this case returns the data from the title column:

sub display_name {
    my $self = shift;
    return $self->title || '';
}

Failing the existence of a display_name method, the plugin attempts to stringify the row object. Using stringification is not recommended, although some people like it. Here is an example of a stringification handler:

use overload '""' => sub {
    my $self = shift;
    return $self->title || '';
}, fallback => 1;

If all else fails the plugin prints the best hint it can to describe the foreign row. This is something approximating the name of the foreign table, the names of the primary keys, and associated values. It's better than stringifying the object the way Perl does, anyway.

Textfields and Textareas

When the plugin creates a web form for adding or editing, it has to choose whether to show a Textfield or Textarea for text-type fields. If you have set a size option in add_columns() within the Schema, and this is less than or equal to 40, a Textfield is used. Otherwise, if the size option is larger than 40 or not set, then an auto-expanding, scrollable Textarea is used.

Column names with spaces

The plugin will handle most tricky names, but you should remember to pass some required extra quoting hints to DBIx::Class when it makes a connection to your database:

# most databases:
{ quote_char => q{`}, name_sep => q{.} }
 
# SQL Server:
{ quote_char => [qw/[ ]/], name_sep => q{.} }

For more information see the DBIx::Class::Storage::DBI manual page or ask on the DBIx::Class mail list.

Database IO filters

Buried within one of the modules in this application are some filters which are applied to data of certain types as it enters or leaves the database. If you find a particular data type is not being rendered correctly, please drop the author a line at the email address below, explaining what you'd like to see instead.

Relocating AutoCRUD to another URL path

If you want to use this application as a plugin with another Catalyst system, it should work fine, but you probably want to serve pages under a different path on your web site. To that end, the plugin by default places its pages under a path part of .../autocrud/. You can change this by adding the following option to your configuration file:

<Plugin::AutoCRUD>
   basepath admin
</Plugin::AutoCRUD>

In the above example, the path .../admin/ will contain the AutoCRUD application, and all generated links in AutoCRUD will also make use of that path. Remember this is added to the base of your Cataylst application which, depending on your web server configuration, might also have a leading path.

To have the links based at the root of your application (which was the default behaviour of CatalystX::ListFramework::Builder, set this variable to an empty string in your configuration:

<Plugin::AutoCRUD>
   basepath ""
</Plugin::AutoCRUD>

Using your own ExtJS libraries

The plugin will use copies of the ExtJS libraries hosted in the CacheFly content delivery network out there on the Internet. Under some circumstances you'll want to use your own hosted copy, for instance if you are serving HTTPS (because browsers will warn about mixed HTTP and HTTPS content).

In which case, you'll need to download the ExtJS Javascript Library (version 2.2+ recommended), from this web page: http://extjs.com/products/extjs/download.php.

Install it to your web server in a location that it is able to serve as static content. Make a note of the path used in a URL to retrieve this content, as it will be needed in the application configuration file, like so:

<Plugin::AutoCRUD>
   extjs2  /static/javascript/extjs-2
</Plugin::AutoCRUD>

Use the extjs2 option as shown above to specify the URL path to the libraries. This will be used in the templates in some way like this:

<script type="text/javascript" src="[% c.config.extjs2 %]/ext-all.js" />

Changing the HTML Character Set

The default HTML charset used by this module is utf-8. If you wish to override this, then set the html_charset parameter, as below:

<Plugin::AutoCRUD>
   html_charset  iso-8859-1
</Plugin::AutoCRUD>

Simple read-only non-JavaScript Frontend

All table views will default to the full-featured ExtJS based frontend. If you would prefer to see a simple read-only non-JavaScript interface, then append /browse to your URL.

This simpler frontend uses HTTP GET only, supports paging and sorting, and will obey any column filtering and renaming as set in your "SITES CONFIGURATION" file.

Overriding built-in Templates

The whole site is built from Perl Template Toolkit templates, and it is possible to override these shipped templates with your own files. This goes for both general files (CSS, top-level TT wrapper) as well as the site files mentioned in the next section.

To add these override paths, include the following directive in your configuration file:

<Plugin::AutoCRUD>
   tt_path /path/to/my/local/templates
</Plugin::AutoCRUD>

This tt_path directive can be included multiple times to set a list of override paths, which will be processed in the order given.

If you want to override any of the CSS used in the app, copy the wrapper.tt template from whichever site you are using, edit, and install in a local tt_path set with this directive.

SITES CONFIGURATION

Another feature borrowed from the original CatalystX::ListFramework is the ability to have multiple views of your data, tailored in various ways. For example you might choose to hide some tables, or columns within tables, rename headings of columns, or disable updates or deletes.

Altering the Default Site

When using this plugin out of the box you're already running within the default site, which unsurprisingly is called default. To override settings in this, create the following configuration stub, and fill it in with any of the options listed below:

<Plugin::AutoCRUD>
   <sites>
       <default>
           # override settings here
       </default>
   </sites>
</Plugin::AutoCRUD>

Configuration Options for Sites

Notes

In the discussion below, the term schema is used to mean the database, and source is used to mean table. These are just more accurate terms for the purposes of this plugin.

In general, when you apply a setting to something at a higher level (say, a schema), it percolates down to the child sections (i.e. the sources). For example, setting delete_allowed no on a schema will prevent records from any source within that from being deleted.

Some of the options are global for a site, others apply to the schema or sources within it. To specify an option for one or the other, use the schema and source names as they appear in the URL path:

<Plugin::AutoCRUD>
   <sites>
       <default>
           # global settings for the site, here
           <myschema>
               # override settings here
               <somesource>
                   # and/or override settings here
               </somesource
           </myschema>
       </default>
   </sites>
</Plugin::AutoCRUD>

Options

update_allowed [ yes* | no ]

This can be applied to either a schema or a source; if applied to a schema it percolates to all the sources, unless the source has a different setting.

The default is to allow updates to be made to existing records. Set this to a value of no to prevent this operation from being permitted. Widgets will also be removed from the user interface so as not to confuse users.

<Plugin::AutoCRUD>
   <sites>
       <default>
           update_allowed no
       </default>
   </sites>
</Plugin::AutoCRUD>

Although this is a separate option from create_allowed, below, be warned! Both creation and editing of records are in fact update actions, so the only way to secure the application from creation and editing of records is to set this option to no.

create_allowed [ yes* | no ]

This can be applied to either a schema or a source; if applied to a schema it percolates to all the sources, unless the source has a different setting.

The default is to allow new records to be created. Set this to a value of no to prevent this operation from being allowed. Widgets will also be removed from the user interface so as not to confuse users.

<Plugin::AutoCRUD>
   <sites>
       <default>
           create_allowed no
       </default>
   </sites>
</Plugin::AutoCRUD>

Important Note: this is purely a cosmetic action, and it is still possible for a user to call the AJAX API published by this application in order to create a new record. The only way truely to prevent record creation is also to set the update_allowed option to no, as shown above.

delete_allowed [ yes* | no ]

This can be applied to either a schema or a source; if applied to a schema it percolates to all the sources, unless the source has a different setting.

The default is to allow deletions of records in the sources. Set this to a value of no to prevent deletions from being allowed. Widgets will also be removed from the user interface so as not to confuse users.

<Plugin::AutoCRUD>
   <sites>
       <default>
           delete_allowed no
       </default>
   </sites>
</Plugin::AutoCRUD>

columns \@column_names

This option achieves two purposes. First, you can re-order the set of columns as they are displayed to the user. Second, by omitting columns from this list you can hide them from the main table views.

Provide a list of the column names (as the data source knows them) to this setting. This option must appear at the source level of your site config hierarchy. In Config::General format, this would look something like:

<Plugin::AutoCRUD>
   <sites>
       <default>
           <myschema>
               <thesource>
                   columns  id
                   columns  title
                   columns  length
               </thesource>
           </myschema>
       </default>
   </sites>
</Plugin::AutoCRUD>

Any columns existing in the source, but not mentioned there, will not be displayed in the main table. They'll still appear in the record edit form, as some fields are required by the database schema so cannot be hidden. Columns will be displayed in the same order that you list them in the configuration.

headings { col => title, ... }

You can alter the title given to any column in the user interface, by providing a hash mapping of column names (as the data source knows them) to titles you wish displayed to the user. This option must appear at the source level of your site config hierarchy. In Config::General format, this would look something like:

<Plugin::AutoCRUD>
   <sites>
       <default>
           <myschema>
               <thesource>
                   <headings>
                       id      Key
                       title   Name
                       length  Time
                   </headings>
               </thesource>
           </myschema>
       </default>
   </sites>
</Plugin::AutoCRUD>

Any columns not included in the hash mapping will use the default title (i.e. what the plugin works out for itself). To hide a column from view, use the columns option, described above.

list_returns [ \@columns | { col => title, ... } ]

This configuration option is DEPRECATED. Please see columns to control which columns are displayed to users and, independently, headings to alter the titles displayed for any columns. The plugin still respects a list_returns configuration setting but will emit a warning to your log that you need to migrate to the new, more flexible, alternatives.

hidden [ yes | no* ]

If you don't want a schema to be offered to the user, or likewise a particular source, then set this option to yes. By default, all schema and sources are shown in the user interface.

<Plugin::AutoCRUD>
   <sites>
       <default>
           <myschema>
               <secretsource>
                   hidden yes
               </secretsource>
           </myschema>
       </default>
   </sites>
</Plugin::AutoCRUD>

This can be applied to either a schema or source; if applied to a schema it overrides all child sources, even if a source has a different setting.

frontend [ full-fat | skinny | ... ]

With this option you can swap out the set of templates used to generate the web front-end, and completely change its look and feel.

Currently you have two choices: either full-fat which is the default and provides the standard full-featured ExtJS frontend, or skinny which is a read-only non-JavaScript alternative supporting listing, paging and sorting only.

Set the frontend in your site config at its top level. Note that you cannot set the frontend on a per-schema or per-source basis, only per-site:

<Plugin::AutoCRUD>
   <sites>
       <default>
           frontend skinny
       </default>
   </sites>
</Plugin::AutoCRUD>

Be aware that setting the frontend to skinny does not restrict create or update access to your database via the AJAX API. For that, you still should set the *_allowed options listed above, as required.

Creating a New Site

You can create a new site by adding it to the sites section of your configuration:

<Plugin::AutoCRUD>
   <sites>
       <mysite>
           # local settings here
       </mysite>
   </sites>
</Plugin::AutoCRUD>

You'll notice that a non-default site is active because the path in your URLs changes to a more RPC-like verbose form, mentioning the site, schema and source:

from this:
.../autocrud/myschema/thesource    # (i.e. site == default)
 
to this:
.../autocrud/site/mysite/schema/myschema/source/thesource

So let's say you've created a dumbed down site for your users which is read-only (i.e. update_allowed no and delete_allowed no), and called the site simplesite in your configuration. You need to give the following URL to users:

.../autocrud/site/simplesite

You could also then place an access control on this path part in your web server (e.g. Apache) which is different from the default site itself.

TROUBLESHOOTING

Foreign keys should be configured with `is_foreign_key`

Any column in your result classes which contains the primary key of another table should have the is_foreign_key => 1 option added to its configuration.

Not doing this will cause the affected column to appear twice in your table, with a (REF) suffix on the heading, rather than once with an (FK) suffix. The technical reason for this, if you are interested, is that is_foreign_key is used by this plugin to differentiate between has_one and belongs_to type relations. The plugin will emit an error level log message if it detects this problem.

If using DBIx::Class::Schema::Loader to generate your Schema, use at least version 0.05 or the most recent development release from CPAN to have this automatically configured for you.

Make sure `belongs_to` follows `add_columns`

Whenver you use belongs_to() in a result class, it must come after any calls to add_column() which affect the foreign key. A situation where this may not be the case is if you add additional column options in a second call to add_column(), after the DO NOT MODIFY THIS OR ANYTHING ABOVE line.

If you do not follow this guideline, then you won't see any related data in the views generated by this plugin. Furthermore, you'll be losing much of the advantage of DBIx::Class, so this plugin will emit an error log level message if it detects this problem.

A better solution is to re-generate your result class using a recent version of DBIx::Class::Schema::Loader from the CPAN (which may be 0.05 or a pre-release).

Optional `belongs_to` relations must have a `join_type`

If you have any belongs_to type relations where the column containing the foreign key can be NULL, it's strongly recommended that you add a join_type parameter to the end of the relevant options to add_columns(), like so:

# in a Book class, the book optionally has an Owner
__PACKAGE__->belongs_to(
    'my_owner',                      # accessor name
    'My::DBIC::Schema::Owner',       # related class
    'owner_id',                      # our FK column (or join condition)
    { join_type => 'LEFT OUTER' }    # attributes
);

If you don't do this, some database records will be missing! The plugin will emit an error level log message if it detects this problem. The technical reason for this, if you are interested, is that DBIx::Class defaults to an INNER join for the belongs_to() relation, but if the column can be null (that is, is_nullable) then you most likely want a LEFT OUTER join.

If using DBIx::Class::Schema::Loader to generate your Schema, use at least version 0.05 or the most recent development release from CPAN to have this automatically configured for you.

Columns with auto-increment data types

For those columns where your database uses an auto-incremented value, add the is_auto_increment => 1 parameter to the options list in add_columns(). This will let the plugin know you don't need to supply a value for new or updated records. The interface will look much better as a result.

If using DBIx::Class::Schema::Loader to generate your Schema, use at least version 0.05 or the most recent development release from CPAN to have this automatically configured for you.

EXAMPLES

The code examples given above in this manual are also supplied in the form of a sample application. You'll find the application itself in the examples/app/ directory of this distribution, and the SQLite3 data source in the examples/sql/ directory.

INSTANT DEMO APPLICATION

If you want to run an instant demo of this module, with minimal configuration, then a simple application for that is shipped with this distribution. For this to work, you must have:

The very latest version of DBIx::Class::Schema::Loader installed on your system (at least 0.05, or the most recent developemnt release from CPAN).
SQLite3 and the accompanying DBD module, if you want to use the shipped demo database.

Go to the examples/sql/ directory of this distribution and run the bootstrap_sqlite.pl perl script. This will create an SQLite file.

Now change to the examples/demo/ directory and start the demo application like so:

demo> perl ./server.pl

Although the instruction at the end of the output says to visit (something like) http://localhost:3000, you must instead visit http://localhost:3000/autocrud (i.e. add /autocrud to the end). Et voila!

To use your own database rather than the SQLite demo, edit examples/demo/demo.conf so that it contains the correct dsn, username, and password for your database. Upon restarting the application you should see your own data source instead.

LIMITATIONS

Single column primary key: There's no support for multiple column primary keys (composite/compound keys). This has saved a lot of time in development because it greatly simplifies the Catalyst and DBIx::Class code.
No two columns in a given table may have the same FK constraint: If you have two columns which both have foreign key constraints to the same table, it's very likely AutoCRUD will not work. Again this is a simplification which speeded the initial development.
Time Zone settings are lost during SELECT/UPDATE: Database fields of types such as (PostgreSQL) timestamp with time zone will be displayed with a date and time picker, but you'll lose the time zone's UTC offset value, sorry. This could be fixed and retained through an improved UI widget, for example.

For the issues above, if you're desperate that the feature be implemented soon, please drop me a line at the address below, because you might be able to buy some of my time for the development.

REQUIREMENTS

Catalyst::Runtime >= 5.70
Catalyst::Model::DBIC::Schema
Catalyst::View::JSON
Catalyst::View::TT
Catalyst::Action::RenderView
MRO::Compat

ACKNOWLEDGEMENTS

Without the initial work on CatalystX::ListFramework by Andrew Payne and Peter Edwards this package would not exist. If you are looking for something like this module but without the dependency on Javascript, please do check out CatalystX::ListFramework.

AUTHOR

Oliver Gorwits <oliver@cpan.org>

COPYRIGHT AND LICENSE

This is free software; you can redistribute it and/or modify it under the same terms as the Perl 5 programming language system itself.

To install Catalyst::Plugin::AutoCRUD, copy and paste the appropriate command in to your terminal.

cpanm

cpanm Catalyst::Plugin::AutoCRUD

CPAN shell

perl -MCPAN -e shell
install Catalyst::Plugin::AutoCRUD

For more information on module installation, please visit the detailed CPAN module installation guide.

	Global
`s`	Focus search bar
`?`	Bring up this help dialog

	GitHub
`g` `p`	Go to pull requests
`g` `i`	go to github issues (only if github is preferred repository)

	POD
`g` `a`	Go to author
`g` `c`	Go to changes
`g` `i`	Go to issues
`g` `d`	Go to dist
`g` `r`	Go to repository/SCM
`g` `s`	Go to source
`g` `b`	Go to file browse

	Search terms
module: (e.g. module:Plugin)
distribution: (e.g. distribution:Dancer auth)
author: (e.g. author:SONGMU Redis)
version: (e.g. version:1.00)

NAME

VERSION

SYNOPSIS

DESCRIPTION

PURPOSE

USAGE

Read Me First

Scenario 1: Plugin to an existing Catalyst App

Scenario 2: Frontend for an existing DBIx::Class::Schema based class

DBIx::Class setup

Scenario 3: Lazy loading a DBIx::Class schema

TIPS AND TRICKS

Update your DBIx::Class Result Classes

Representing related records

Textfields and Textareas

Column names with spaces

Database IO filters

Relocating AutoCRUD to another URL path

Using your own ExtJS libraries

Changing the HTML Character Set

Simple read-only non-JavaScript Frontend

Overriding built-in Templates

SITES CONFIGURATION

Altering the Default Site

Configuration Options for Sites

Notes

Options

Creating a New Site

TROUBLESHOOTING

Foreign keys should be configured with is_foreign_key

Make sure belongs_to follows add_columns

Optional belongs_to relations must have a join_type

Columns with auto-increment data types

EXAMPLES

INSTANT DEMO APPLICATION

LIMITATIONS

REQUIREMENTS

SEE ALSO

ACKNOWLEDGEMENTS

AUTHOR

COPYRIGHT AND LICENSE

Module Install Instructions

Scenario 2: Frontend for an existing `DBIx::Class::Schema` based class

`DBIx::Class` setup

Scenario 3: Lazy loading a `DBIx::Class` schema

Update your `DBIx::Class` Result Classes

Foreign keys should be configured with `is_foreign_key`

Make sure `belongs_to` follows `add_columns`

Optional `belongs_to` relations must have a `join_type`