|Did you know ...||Search Documentation:|
|Pack weblog -- docs/form redesign discussion.txt|
aw, no prolog at languages
yup, we should sue!
so, wahts yer thoughts on form?
well, I'm not sure if we should have separate modules; one for simple forms and one for validated
maybe just some extra arguments for the validated forms (
form(..., ..., [validate(....)]))
imagine the dev process
somebody makes a form
later they realize they want it to validate
they shouldn't have to undo/change what's done
so yah, some 'extra arguments'
How do we specify the validation?
and what happens if it doesn't validate?
is it server or client side? what if the validation can only be done server side?
well, some things (unique name) can only be done server side.
also, the server should check if the fields are valid even if the page has validation client-side
in case the user bypass it
so for starters maybe we should focus on just server side
well, i think it will be nice to be able to add "validation specification" at the same place you declare the form field
what about error message?
like, say this field has to be non null
so if you submit that way, should be
name must not be blank name [___________]
how's that specified? well, if we have some default validation specifications (like, notnull, alnum...) we can have default error messages but the message text always needs customized
otherwise, it should be specified when you declare the validation ok, that's reasonable (well, they can also customize the error in default) some people might want to make the err message someplace else (eg all messages at top of form) hmmm... create some sort of error event and let one process collect them and display them based on user specification perhaps? kinda messy though
Yes - how about a dcg for the error message - expands to nothing if the error's not activated? Not sure how that would enable the user to display the errors at the top :/ oh, I think I get it (i'm sencing race condition danger in our chatting method xD) I like the race condition XcD
form(action=formlanding.php, [ \error_msg([for=name, violation=foo], 'You must supply a name'), \error_msg([for=name, violation=bar], 'must not contain spaces'), p(\form_field(input([type=text, name=name], )), [foo(length>=3), bar(not(contains(' ')))]), ... ])
something like that?
hm, that raises the question; what if we want different messages for different kind of validation violations? for example, "You must supply a name" "Name taken" "Name should not contain spaces"
hmmm there is also this possibility:
yes, that's nice
\error_msg([for=name, length>=3], 'You must supply a name'),
\error_msg([for=name, not(contains(' '))], 'must not contain spaces'),
p(\form_field(input([type=text, name=name], ))),
and, for the record, until we got on the issue of different types of violations, we were following my existing code - which is some sign we're not totally insane
(my thing was great to write, it just is hard to debug)
we are prolog programmers. I think one could say that we are insane by default :b
True. OK, so how do you handle the decision to reload the form or proceed to the landing page?
A gotcha I ran into was that first time you visit form it has error messages - like, name should only have error msg shown in above if you've already submitted form once
Well, the dirty way to fix is to claim that they are not error messages, just guidelines for the user :b but that's bad. so hmm...
yah, I handled it by looking to see if the parms were there if the name parm exists and is null then err message if no parms, we're on first load
just warning you if you're going to try to do something same, to handle this case
maybe some hidden form field?
OK, so how do you handle the decision to reload the form or proceed to the landing page?
so, if we are doing everything server side, we can validate the fields; check all conditions, no short-circuiting: I once was trying to register and didn't get all the error messages the first time and it was really annoying
I agree - should show all the error messages, not stop at first
I'm asking what my, as a programmer using your code, method is to land the user on the right page?
hm. you might want to return to the last page after, for example, registering. hmmmm
oh, good point - we should support that also, the app programmer needs to know if the form's valid, my code generates the HTML twice. That's so it can handle error messages before or after the form field. But your way of putting the validation in the error message is better. and eliminates that to boot but still, we need some way of having the result of pressing 'submit' land you on the right page, either the landing page, or back on the form if you've made an error.
and maybe erase some fields, like password (not sure why they do that) and CAPTCHA
well, captcha we handle as a field type the session remembers what we gave as challenge
you erase password fields so they don't get transmitted in plaintext in the form.
oh my method, and this is the sucky part of mine, is to have a wrapper on the handlers
handler('/some/form', myform_handler , ).
validated_form(Request, actual_form_handler, landing_page_handler).
it calls actual_form_handler twice - needs to for the 2 pass compiler issue your put validation in error message thing eliminates if it's valid, it then calls the landing page handler
hm, couldn't we have some AJAX and perform the validation asynchronously - and update the page?
we could, but I think we should handle the old fashioned way since we have to handle it anyway
it'd be very very nifty if you could just set a flag somewhere and it does all this by ajax with client side verify. But I think the default has to be the old 1990's style
well, we're compiling, essentially - it's a preprocessor on Jan's termerized HTML. actually, that's the bad part of mine - it's really hard to figure out hwat the hecks going on through the indirection of validated_form
so what's yer thoughts on how to invoke the validation and end up on the right URI ??
hmmmm. not sure. cannot think of something good :/
sorry, need the trick:
html([.... form ...]),
We redirect output into a string and capture the form. If we're valid when we reach end_validated_form, we discard the string and call the landing page.
the 'trick' is that you have to hang onto a boolean that is set by start_validated_form and might be reset by an err_message DCG.
So you need one of the stateful dodges - thread local assert or something
or something like setvar right?
yes nb_setval or something
tbd decided what
remember the server's multithread, you gotta do it per thread
this looks much better than my existing code
hehe. Do you think it's a good point to stop for a bit? I gotta eat XD
I think this is a good design Are you willing to try implementing it?
sure, I'll try!
CVool! Yay - I think between us we fixed the issues with existing, this is gonna be great. I'm going to spend some time thinking about other parts of design.
:) cya saves chat yup