NAME
Data::MuForm::Manual::FormHandlerDiff - Changes from FormHandler to MuForm
VERSION
version 0.04
DESCRIPTION
This documents the differences between FormHandler and MuForm, in order to assist converting forms or deciding which package to use.
Since this is an entirely new package I have taken the opportunity to make all of the changes and improvements that you can't do when you have actual users, including renaming things, removing features, changing behavior, and removing cruft. A very large part of the core FormHandler code and functionality has stood the test of time and use very well, and has been moved to MuForm with only minor changes.
Form construction is substantially faster in MuForm, for a number of reasons: Moo instead of Moose, fewer attributes, no results, and dropped some complicated only marginally helpful features. Processing is somewhat faster. Rendering takes about the same amount of time (most of the extra rendering related time in FH is actually at form construction time).
Note: this is still a work in process, so please contact me if your favorite feature has gone missing. Some features may be supplied via roles.
In your MuForm classes do;
use Moo;
use Data::MuForm::Meta;
extends 'Data::MuForm';instead of:
use HTML::FormHandler::Moose;
extends 'HTML::FormHandler';In your MuForm roles do:
use Moo::Role;
use Data::MuForm::Meta;instead of:
use HTML::FormHandler::Moose::Role;The 'result' objects have been removed because there was too much overhead and complication for low payout. Some rendering tasks were practically impossible to support using the results.
The 'item', 'item_id', and 'item_class' have been renamed to 'model', 'model_id', 'model_class'. The 'model' is now cleared on process like the other attributes.
When a DBIC row created with 'new_result' is used as a model, it will not be used for defaults unless explicitly overridden in the form with a 'use_model_for_defaults' sub.
The 'init_object' has been renamed to 'init_values'.
Note: the rendering is entirely different, so much that there's no point in documenting the individual differences here. No more widgets. I found that the rendering of form elements hardly changed at all, and that trying to have one wrapper method that handled everything ended up with a large mess. Attributes that were rendering specific have been mostly moved into a render_args hashref that will be passed to a renderer, in order to allow more flexibility and reduce the overhead of form construction. I attempted to make it easier to render in templates with 'render_element' while still allowing full-automatic rendering. There are 'layouts' for some form controls.
In particular, the defaults for wrapping fields have changed, for compound and repeatable fields, etc. I advise you to make a test case, like the ones in the t/render directory, and ensure that the output matches, if you are converting from FormHandler to MuForm. Compounds and repeatables have wrappers and labels by default. Repeatable instances do not have labels but do have wrappers.
See Data::MuForm::Manual::Rendering for more rendering information.
Renamed 'init_contains' to 'init_instance' which is more accurate.
Set options attributes directly in the 'options' hashrefs, instead of in a separate 'attributes' hashref, i.e. { value => 1, label => 'one', class => 'xxx' }
No 'has_block', since that's purely a rendering thing. There will be some way of doing something similar in the renderer.
Localization now uses a gettext style .po file and a gettext type localizer, with named parameters. Users can supply their own localizer as long as it can use the translation files and supports the calls that the internal add_error* methods use. Gettext has more complicated methods for localizing, so a few additional add_error* methods have been added, but shouldn't be needed very often.
Added 'check' method for validation, which doesn't do 'update_model', for non-form type validation.
Added 'data' alias for 'params', for better naming when doing simple validation
Renamed 'field_name_space' to 'field_namespace'. ('widget_name_space' is gone)
Standardized returns from attributes to arrayref, instead of array. If an array return is available it will be in the 'all_' version. The important ones are 'options' and 'fields'. A field 'options' returns an arrayref of options. 'all_options' returns an array. The 'fields' attribute returns an arrayref of fields, 'all_fields' returns an array.
The trim method is a separate coderef and is not added to the apply actions. Needs to be defined as a simple coderef instead of an apply-style action.
The 'render_filter_method' has been removed. The 'html_filter' is defined in the Renderer. If you want a different filter, create your own renderer (recommended anyway) and replace it.
The active => [] and inactive => [] no longer have a permanent build (on new) effect. Do in your own form code if you want this. They still work the same on ->process. The 'active' and 'inactive' flags are implemented slightly differently, but shouldn't affect field definitions.
The 'posted' flag has been renamed to 'submitted', because you also need on a 'get'. Also a params hashref with no field name keys will not need the submitted => 0 flag.
You no longer need to use $field->_set_value($value) or $field->_set_input($input). Just use $field->value($value) and $field->input($input)
Removed 'update_subfields'. The code was complicated and not worth the overhead. If you want to do something like this, see the cookbook.
Removed 'use_fields_for_input_without_param'. Happens normally now, since the default is to process all active fields unless the 'skip_fields_without_input' flag is set.
Removed flag 'use_init_obj_when_no_accessor_in_item'. This should happen automatically now.
Removed flag 'use_init_obj_over_item'. Just use init_values in your process call instead. Also a 'new_result' model will not be used for defaults now, so it's not as likely to be necessary.
Removed flag 'use_defaults_over_obj'. Use transform_default_to_value instead.
Removed 'no_update' flag on form. Use 'check' method instead.
Removed 'update_field_list'. Can be implemented easily via user code if needed.
Removed 'defaults' on 'process'. Use init_values instead or implement via user code.
The inflate/deflate transforms have been renamed: 'inflate_method' = 'transform_input_to_value', 'deflate_method' = 'tranform_value_to_fif', 'inflate_default_method' = 'transform_default_to_value', 'deflate_value_method' = 'transform_value_after_validate'. Added new transform: 'transform_param_to_input'.
Removed 'deflation'. Many other ways of doing this.
Removed 'missing'. I never used this. Did anybody?
Various methods that could be set in a field definition have been moved into a 'methods' hashref: 'build_id', 'build_options', 'build_label', 'validate', 'default'.
TextArea field renamed to Textarea for consistency.
A number of trivial fields are no longer supplied by the distribution. If you need these fields it should be pretty easy to create your own. Some fields not supplied: Hour, Minute, Month, MonthDay, MonthName, Second, Weekday, Year, BoolSelect, DateMDY, PosInteger, Duration, PasswordConf, IntRange. The DateTime field has been renamed to 'CompoundDateTime'.
Renamed html_prefix to field_prefix and it should be set to the string used to prefix fields.
Renamed field 'noupdate' to 'no_update'.
Added 'build_field_id' as a possible form method to build field ids.
Renamed push_errors and push_form_errors to push_error and push_form_error.
Internal method 'validate_field' (sometimes used for field tests) renamed to 'field_validate'
Removed internal method 'base_validate', added 'normalize_input' instead (for handling multiple/non-multiple, etc)
Renamed 'html5_type_attr' to 'html5_input_type'.
AUTHOR
Gerda Shank
COPYRIGHT AND LICENSE
This software is copyright (c) 2017 by Gerda Shank.
This is free software; you can redistribute it and/or modify it under the same terms as the Perl 5 programming language system itself.