pirsquared17 / activescaffold Goto Github PK

I've been playing with this.

> def #{column.name}_column(column_value)
>   # column_value is the rendered value of the column that we would
> normally display: h(format_column(value))
> end

This version does not seem useful since the column value is already
preformatted and has no access to the full record.  You mention that
it might be useful for wrapping the value in a link but usually the
link will require some data from the record as parameters.  I would
suggest removing this version and just relying on (column_value,
record).  It actually makes development simpler since I don't have to
worry about which method version should I use.

-Mike

use flash[:error|:warning|:info] when respond_to type.html

Remove hardcoded colors from Rico.Corner.round(). Investigate using other
libraries that don't need hardcoded colors (and handle tables in IE).

right now column totalling is on a per-page basis. it should instead report
the column total from all records in the result set (from all pages).

Now that method sorting uses Ruby's array#sort_by, it doesn't handle
inconsistent data as well. It fails when comparing nil to fixnum, for
instance. Not sure how best to resolve ... type-casting is tricky because
it doesn't know what the target type should be.

Check Scott's sorting method again for clues.

had an idea to replace dhtml_history.js. if we could reliably intercept the
back button (or other history traversal events), we could move the client
history server-side. this would provide a broader and more reliable history
solution (the current one is FF/IE only, i believe).

if session storage becomes a concern, we could limit a person's history to
a number of pages, say, five.

the spinner appears in the first column, no matter which column you're
sorting. reported by scott.

would allow users to import records from csv or xml

What steps will reproduce the problem?
1. Click "Show"
2. Click "X" to close
3.

What is the expected output? What do you see instead?
I expect the show view to replace the current view. Instead it appears to
the right. Then, I expect the X to close the show view. Instead it reloads
the page.

I can fix these two bugs by adding "if (window.opera) throw('use
ContentFromAnonymousTable')" to Abstract.Insertion.initialize (line 1628 in
Prototype 1.5.0). But then Opera fails to refresh the original row correctly.

I don't want to start patching Prototype until I can solve all the issues
together.

We either need upgrade our rounded corner JS library to one that will work
cross browser without requiring the colors to be hard coded.

1. view a list with no entries
2. add an entry

the nice little note about not finding any entries is still there.

enter issues for opera support into tracker

would allow user to build field/operator/value clauses and join them
together with and/or operators.

searching is case-sensitive in postgresql. currently the search term is
downcase'd, but the field is not.

consider: if we downcase the field, will that ruin the indexing?

it can be useful to customize the css class of a <tr> or a <td> based on
properties of the record. for instance, highlighting a row (or cell) in red
if it contains a negative value (consider a financial data table).

we could be minimal and only support customized css classes for the <tr> or
the <td>. if we only supported customized css on the <tr>, then people
could style individual <td> based on the standard css class. if we only
supported <td>, then people could style the <tr> by styling each <td>.

options for implementation:
  1) conventional helper method
  2) a proc in the list configuration

would allow user to request and download result set in XML or CSV format.
different from the RESTful API because it allows people using normal
browsers to click-and-save.

If someone wants to extend the JavaScript response for a response generated
by render :update, they need to override the entire method. If we move
entirely to RJS files no matter how simple the response, people can
override the template independently.

And yes, I know, I'm the one that put the render :update calls in there in
the first place.

* http://mir.aculo.us/2005/10/03/ruby-on-rails-i18n-revisited
* Spanish, German, French, English (official files)
* global configuration setting (like theme)
* override option

comb through search for sql injection opportunities

need to do a once-over in internet explorer (6 and 7) and enter issues in
tracker

would allow user to build a list of field/operator/value clauses (multiple
clauses per column)

What steps will reproduce the problem?
1. Open live search
2. Optionally type something in search box
3. Press enter with focus in search box

What is the expected output? What do you see instead?
I expect the search to process. Instead I get an exception report about
interning an empty string.

Use form/button for delete so that we can always make sure it’s using POST

Run XHTML through a validator and enter issues into tracker.

Creating records in a Nested scaffold lose their parent's association.

It is being removed on line 43 of

Actions::Base::ClassMethods

action links could open in (javascript) popups. popup size would be
configurable.

all of the do_#{action} methods should be protected, so that they can't be
accessed via url (should get a 404, not a 500).

find any issues related to RESTful compliance and enter into tracker

record count doesn't update after create/delete

support ActiveRecord Annotations when available. use annotations to improve
the form (widget selection, validation), list (special rendering, e.g.
phone number formatting or image thumbnail display), and search (in case
people search using the formatted value instead of the database value).

momentous.

while you do that let me voice my desires for refactoring around _nested.rhtml

for sub forms I currently have a (crude) way of finding a scaffold config given 
the model class, I feel 
like a lot of the work in _nested should be integrated in with that helper 
method

we should be able to if given a model ask helper to give us a scaffold config 
for that model and get 
the controller name if necessary

we should also use render :active_scaffold instead of render_component there

and perhaps render :active_scaffold can take a model class and go off and find 
the controller on it's 
own, I dunno but we need to push that into some helpers and out of this 
particular view
okay... just wanted to get that into the transcript :)

By default, nested scaffolds show the reverse association column (e.g. if
the UserGroups scaffold displayed a nested Users scaffold, then the Users
scaffold would still display info about the related UserGroup). This column
can't be removed because then it triggers bugs like reported in issue #43.

If the reverse association columns were automagically hidden, then we could
avoid poking holes in the security filters and still provide a usable UI.

column totals need to update on create, update, and delete

We need to ship with a basic search model that works for people without
JavaScript.

transparent pngs lose transparency in ie 6.0.

When a controller is sub-classed and ActiveScaffold is configured on the
parent, the child doesn't work. See forum thread for more.

Searching uses a LIKE '%?%' syntax, which is not indexable. Consider using
LIKE '?%' instead.

* Change “themes” into “frontends” and create new stylistic concept of 
themes
* Have assets overwrite based on frontend
* Add note to assets directory that it is dynamically generated
* simple configuration for frontends: color, form style, etc.

sorting needs use a whitelist so it doesn't try to use unknown columns

For a while I have been ignoring a problem in which the formatting of
the totals row intermittently goes AWOL.

To recreate the problem I need to start with a record set with an odd
number of records showing.  The problem can then be re-created by
opening and cancelling the update form.  Before opening the update
form Firebug tells me that the the total row is under the influence of
the "ajax-scaffold-totals" CSS class.   However, after opening and
closing the form the total row is under the influence of the non-
existent "ajax-scaffold-totals even" class.

This none existent class is being over-ridden by the ".ajax-scaffold
tr.even td" class... thus the formatting problems.

Refreshing the page corrects the situation and simply creating a "ajax-
scaffold-totals even" class in the CSS does not seem to sort the
problem. And there is no problem with an even number of records!!

Can anyone else reproduce this effect?  If so.. reported as a
problem.  If not, any tips on CSS entries that may sort my problem
gratefully received.

Neil

----------

The background and the border are affected but I have hidden border
problem by setting it to the same as "ajax-scaffold tr.even td" class.

LiveSearch should not support html interaction. It just shouldn't even
submit (or open) in html mode.

would allow the user to specify search terms on a per-column basis

would delete all records matching current search

after editing/creating/deleting associations in a nested scaffold, the
related association column does not update

use relative font-size specifications in stylesheet, and only specify an
absolute size on the base class (div.active-scaffold). this lets a designer
change one value to scale the scaffold to the page.

when reworking nested (embedded) scaffolds to use constraints, i lost the
feature whereby the constraints customize the scaffold label.

difficulties in reimplementing this:
* localization (possible solution: use parenthesis instead of "for")
* which constraints to use? just association-based ones?

maybe we actually want to customize the scaffold label as part of the
'render :active_scaffold' call. then it's in the dev's hands, but we can
still make it slick for nested scaffolds.

If a column has been formatted so that it no longer parses as a number
(e.g. has a currency symbol in front), then totalling will not work on that
column.

would allow user to update specified fields to specified values for all
records matching current search terms

...probably in postgresql

the LIKE operator doesn't do well with datetime or boolean fields.