Generic Multiple Value Handler for CCK Fields

This is my project proposal for the Summer of Code 2007. Better multiple value handling in CCK is needed and could be useful to a lot of people, so I am interested in having your opinion before I officially submit this project proposal to Google.

Motivation

The current implementation lets individual CCK field modules handle multiple values themselves. This approach is simple but it limits the flexibility of CCK content types in various ways. Issues with multiple values are very common in the forums and on the CCK issues list, but the answers received are never straightforward and clean.

The proposed project is a module that would set the bases for a new method for handling multiple values of CCK fields that would be more unified and flexible than the current implementation. The main goal is to be able to group multiple values together so that the database schema reflects the logical organization of the data.

A practical example is easier to understand than an abstract description, so suppose I want to create a recipe content type using CCK. For the ingredients, I need a list of items containing: the description of the ingredient (text), the quantity (number) and the unit (select: grams, liters, etc). Logically, the database structure should be something like:

but currently, if I create fields for ingredient names, quantities and units, all of them having multiple values, I get three tables of the form:

Having the data stored this way in the database makes it difficult to update, sort or view it in an efficient way while keeping relationships between row elements consistent.

Several workarounds currently exists, fieldgroup_table, subform or editview just to mention a few. I recognize the effort put in these modules, but fieldgroup_table has unsolvable input validation issues, and editview and subform require to create nodes for sub-elements, which is not always appropriate (for example in the case mentioned above, 10ml of lemon juice does not mean anything by itself). What we need is a solid module that reflects the logical structure of the data from the user interface to the database.

General Idea of the Implementation

The "multiple values" attribute of individual fields is replaced by a new field type that acts as a container. Let's refer to this new field type as a "multiset" field. Any field type could be created inside a multiset.

In addition to the fields it contains, a multiset field would also have settings for the input method. Possible widgets include:

• a list of fields similar to the current input widget for multiple values. The number of empty fields presented to the user on node creation and modification is customizable.
• a list of fields with a link or button to add a row using JavaScript (AJAX is not required since all elements added follow the same template). A link or button to delete a row should also be available on the side, since setting many fields to their empty value can be cumbersome, especially when using multiple select widgets (like the date module).
• a predefined set of fields, each of which is associated with a key value in a similar manner as the inventory module (link). Depending on the implementation details, this kind of multiple values might need to be handled in a separate module using the same common functions with only some slight differences.

Setting the 'required' attribute of a contained field would then indicate that any non-empty row needs to have a value for that field.

If the data is organized logically at the database level, it will then be easier to implement different formatters for multiple-valued fields.

Planned Issues

Since CCK has rapidly become an essential contributed module for many drupal-based websites, compatibility of the new module with current modules is crucial. By ignoring the 'multiple' attribute of the fields inside a multiset field, the new module can use any CCK field without needing to make changes to existing modules.

There also needs to be a script or a separate module that make it possible to update current data to the new structure by converting fields with multiple values into multisets containing the original fields.

This multiset module provides powerful but abstract functionalities, so having a clear and meaningful user interface is important in order not to confuse inexperienced users.

Long term

When this module becomes part of the core CCK modules, the creation of new CCK field types will be made easier because field module developers will not need to write (redundant) code to handle multiple values.

Credit

The motivation came from working on a few Drupal-based websites that gave me trouble with multiple values handling in CCK. However, the idea to step apart from CCK's current multiple value handling at the database level was inspired by the Inventory module.

I'd like to thank Robert Douglass for his useful suggestions to clarify the goals and purpose of this project, as well as the many Drupal developers without whom this project wouldn't be possible.

About Me

I'm a student in mathematics and computer science, and I do web development as a part-time student job.

After having worked as an intern or a part-time employee in different software companies, I have realized that open-source does make a difference. I believe in the values it represents and I hope that the Summer of Code experience can be a first step for me towards a professional career in the open source world.

99% of the software I use is open-sourced, whether I'm writing a paper, doing development work, recording music with my band or simply reading my e-mail, and I have convinced people around me to do the switch too (my girlfriend now runs kubuntu on her laptop and my college's website will be drupal-based as of next fall!).

Considering how much I have gained from Drupal and open-source software in general, I'm extremely excited at the thought of spending the summer making a contribution to this wonderful project.