error messages should go. (It now occurs to me that maybe the form generation should get the errors from the validation instead of htmlfill). The form generator might be a person (totally customized and easily tweakable), or a template (e.g., made with Cheetah or ZPT), or an HTML-generating model. All that matters is that it creates a form; and you can change the technique without changing the rest of the system.
Then htmlfill takes it all and with its knowledge of HTML forms puts it together. Note that this is different from most templating languages. Some languages know about XML or HTML (e.g., ZPT) and some know only about text (e.g., Cheetah); in this case we are very specifically looking for input elements (and other similar elements), and saving all the extra annotation that might be necessary.
(Note that I have a validation package, and I have htmlfill, but a form generator is left up to the reader.)
The contrasting technique (the one that formencode.htmlview uses) is to do all of these at once -- you prepare the model, the defaults, maybe a previous request, the errors and pass them all in. Then you go through the widgets and they each render themselves, enclosed in some wrapper. I won't argue for why this doesn't work well -- if you've tried it (I have, a couple times) you probably know its limitations and complexities. This constrasting design doesn't even have "widgets", just plain old HTML. Good riddance!
Subway should really use it, it is the cleanest form toolkit approach ever...# Ksenia
This is an interesting little problem that i had to face also a few weeks ago when I wrote part of a web application framework. I took a really simple approach which works well with my framework (I like simple things, this is one of the reasons I like Python) and I will describe it below.
BTW, I find your log entry a little bit difficult to understand for someone that is not familiar with your formkit code, and I think the problem is interesting enough that it deserves more details. It would be great if you could write your thoughts in a self-contained document with clearer definitions of the problem to be solved, etc. i.e. not just a blog entry, but a design doc. Just an idea.
Here is how I solved it in my framework (this is tested code and running in a web app).
In summary, this problem can consists of separate components which interact in well-defined ways, here are the definitions of my components:
- form definition: listing the names, types, constraints and labels of the forms;
- rendering a form: this involves taking the form definition, a set of initial values and generating an output that represents the corresponding HTML (or directly outputting HTML text);
- error markup: taking potential errors from a previous request adding markup to indicate errors near the corresponding input fields;
- validation: taking the values sent from the submit of a form and validating the constraints on the widgets, potentially returning the user to the form if there are errors;
- conversion: converting the string values from the form submit to Python data types.
Before I go on, I must mention that in my framework I do not output HTML as I go. Instead, I build a tree of HTML tags in memory (using a special, really really simple library that I built for this purpose (htmlout), I can provide it if you want); this allows me to manipulate various parts of my document in any order before rendering it, remove stuff, change attributes, classes, etc.
Using this method, I build a custom Python class to define a template for each of my page layouts. These classes have methods for each layout component, e.g. add_sidebar_paragraph(text). This is a good way to force yourself to "design" most of the layout upfront.
I very much prefer this approach to any of the billion text templating systems because I can change parts of my documents in any order, and I can add many "smarts" to my template layouts. It's code, it's dynamic, rather than blobs of text to be pasted together. Note that this can only work out if I don't collaborate with artists (which is my case at the moment), i.e. you have to write Python code to generate the HTML. I suppose if I did work with designers I would have to hook a templating system in.
This might have had an impact on the design of my system, but I think most of the ideas below still apply with a usual templating system.
I defined a library with a "widget" object for each type of entry (string, multi-line text, menus, radiobuttons, file upload, date, etc.). In that library, there is a "form" class that acts as a container of these widgets. It can return the list of labels, names, widgets, and can parse a dict with data coming in from a request. If some validation fails it raises an exception. After parsing, a new dict is returned with the string input values converted into Python native values (e.g. a datetime object, a unicode string, a file buffer, etc.). This library contains no rendering code.
I wrote a simple convenience method that calls the form parsing and catches the form error exception; if an exception is raised, it sets a status message (in per-session dat) for the next request to render (a message to be written to the user), and serializes the parsed input values (a dict) AND a list of error fieldnames and error messages (a dict also) into per-session data in a database. Then I redirect to the render request(*).
Conversion occurs at the same time as validation. This is for efficiency: oftentimes validation requires conversion. I have not seen any problem with this approach yet, since converting is always done at the same time as validation. This is the first thing I do in a form handler method (after authentication checks).
- (*) Note: the redirect is not really necessary, I suppose I
- could save a request by calling the handler directly with the values dict and error dict.
I have separate renderer objects that implement rendering using single-dispatch on the form widgets (i.e. def render_StringField, def render_FileUpload, etc.). I have three:
one renderer that can generate an entire form generically; useful for debugging, but if you build fancy website you almost always need to organize the inputs nicely and customize in a way that usually cannot be figured out generically; so most of the time I use...
another renderer that generate mini-tables for a sublist of widgets that are defined in the form. I call something like:section.append( H3(_("Travel Parameters")), form_renderer.rendertable( form, values, fields=['departure', 'passengers'] ) )
a "display" renderer that generates tables for display purposes only, using the form definition, and using again a sublist of widgets/values to render. This is useful because oftentimes you want to display the data and much of the information about how to display it is already encapsulated in the form.
And since the renderers and the form definitions are not dependent on each other, it is easy to do that. I suppose I could add other types of renderer if I can find other uses for the form definition.
In all cases, I pass in the values dict that is read from the per-session data and the renderer code knows how to undo the conversion and fill in the values. There is also a phase for the widgets to "prepare" the input values before the renderer uses them. I use this to undo some of the conversion.
A note about internationalization: if the form definition occurs when loading the modules, the renderer has to know to gettext() the labels before rendering them.
When I want to "render" errors, the handler code passes the HTML form tree generated by the code above (including custom HTML formatting that is added to make the form look good, sectioning, fieldsets, etc) to an error renderer method.
That renderer runs down the tree and finds the first input that corresponds to an error field and inserts appropriate marker HTML to indicate to the user where the errors are located and adds some CSS classes to the HTML inputs. I think this is equivalent to the htmlfill library, but it might be more efficient because I don't need to "parse" the text. This happens before flattening the form into text.
Thanks for the (very extensive!) comment.
This sounds a lot like FunFormKit, which was an earlier library I wrote; maybe it's a natural progression. In that library the components were largely how you describe, though error markup happened at the same time as rendering. One major issue was how to deal with dynamic forms. FFK had ways to do this, but it was complex and non-intuitive, even for very straight-forward forms of dynamicism (like dynamic select boxes).
I wrote a small library to generate HTML, mostly for me so I could get rid of a lot of HTML-related logic. The layout was done similarly, but I now feel it's not sufficient -- not just when you are working with a designer, but for that last 10% of a project where you start caring about the little things. I would feel a need to make changes to the layout system for every new form once I got to the last 10%, and that's no good -- the form library should be stable.
For signaling errors to a user, you definitely shouldn't do a redirect or store the errors in a session object. It's easier just to redisplay the form with the errors inline, and make them re-POST the data. One trap to avoid that FunFormKit fell into is overdesigning this process -- let the application call into your library and do the basic control, don't try to hijack the request when an error occurs. That gets way to complicated for no real gain.
Interesting. Could you write a short architectural overview of the new system? I'm really interested to find out about the differences in high-level design.
About the HTML generation library: the lib I wrote is a simple mapping to xhtml, i.e. it can generate any xhtml code that you could write in text, I don't find any limitations. It's really just like an XML tree but the various tags are defined as classes themselves, that's all.
For errors: you're right, and indeed I'm not "keeping" the form data in the session, just using the session as a temporary storage to communicate it to the render request for re-rendering the form (since they are separate requests, different children might handle it). The user does re-post all the data.
Oh, I think I get the misunderstanding, I think I forgot to mention that I wrote my code so that requests "just render" and "just handle" are separate urls. i.e. /contact_edit renders the form, /contact_edit_hndl handlers the submit data, and then either redirects back to /contact_edit or to some other page (in case of success). I was debating for a while whether this was a good approach (my background is in desktop apps) and I pretty much like the separation. I don't like for the requests that "just render" to check and "maybe" handle the submit data.
I love FormEncode--it's be best concept of a form validation tookit I've ever seen. However, in all of the tutorials and examples I've seen the actual form markup is maintained as a string that is not part of the page template (it gets inserted into the template later). I don't like that. I want to have all my form markup reside in the template (I'm using SimpleTAL), and process the template with htmlfill after the template is expanded. I realize that there may be a limitation of one form per page (or at least every input tag must be uniquely named on each page).
Comment: htmlfill must happen after the template is expanded because sometimes input tags are generated by the templating system. If htmlfill happened first these tags would not exist.
Bottom line: I'm running into problems with SimpleTAL stripping the markup out of my <form:error name="fieldName" /> tags--after SimpleTAL expansion they look like this: <form>. That won't do. Is there a workaround or planned solution to this problem?# anonymous
Sorry, forgot to include my name...
OK. I found the solution. Duh ... declare "form" as an XML namespace. Like this:
<html xmlns="http://www.w3.org/1999/xhtml" xmlns:form="http://www.formencode.org">
That's all it took. Thanks for providing a space for me to think out loud Ian.