Age | Commit message (Collapse) | Author | Files | Lines |
|
This also adjusts the method for comparing filters, allowing a chart to
define a "filterCompare" method that will be used to compare filters if
available.
This finally removes the date chart special case for setting filters.
Array instance checking will now use foo instanceof Array instead of
typeof(foo) == "array", which was unreliable (sometimes it's "object").
Trellis chart brush containers will now redraw the brush when set
externally.
renderBrush no longer takes brushG as a parameter. I'm not sure what I
was thinking when I made it so.
The deselected area will not yet fade.
The trellis chart still requires substantial refactoring and consistency
fixes; I'm not sure that the parameter specifications are currently
correct.
|
|
Before, if you were quick or your connection was slow, you could select the
files before mv.init registers the callback.
This left things stuck, either requiring another set of files to be
selected, or a refresh of the page.
This closes #18.
|
|
|
|
mv.loader.use takes an array of URLs, which are fetched via d3.xhr.
This also restructures mv.loader to refactor out some shared functionality
between mv.loader.use and the file selector change events.
mv.loader could still use some cleanup.
http://css-tricks.com/examples/ButtonMaker/ makes a pretty neat css button.
Hopefully, it's visually attractive enough so people in a hurry know where
to go. :)
|
|
|
|
The inner loader closure was redundant. Apparently I really like closures
during late night coding.
main.js should no longer have to deal with raw loading events; only UI
specific callbacks.
|
|
Also remove a stray curly brace
|
|
.zip must be at the end of the file name.
Progress bars are not fully descriptive in this mode yet
This also alters the mv.loader interface.
The onprogress callback now takes current and total.
The each callback is now passed an after parameter, to be called when it's
done.
|
|
|
|
This should all be done by delta updates.
d3 joins, while fun, aren't really the tool for the job here. :)
|
|
|
|
|
|
|
|
Previously, you could still click on the (fully transparent) pieces.
|
|
|
|
This should hopefully make things apply more cleanly for reconnections and
fallbacks.
|
|
This is by no means even close to comprehensive (legacy compatability for
determining browser width is a total mess in and of itself).
However, this is enough for manavis to work on recent versions of Firefox.
It's not quite as fast as chromium, but it is entirely workable.
Older versions aren't really worth testing, as the entire application is very
js intensive.
|
|
|
|
Active filters will only be applied from users in the same channel.
This closes #14.
|
|
It was previously very hard (impossible?) to select upper values.
|
|
|
|
Also filter KILLXP by default.
This closes #16.
|
|
|
|
|
|
Sort the pie chart the other way.
Still doesn't feel right. Perhaps a bar chart would be better.
|
|
|
|
The new parsing system is robust enough to eliminate its persistent need
|
|
|
|
No elegant solution presents itself immediately. This isn't too bad, but
should be replaced - while more specialised charts could be added to the chart
renderlet, it's fragile and matters become problematic if more modules require
the renderlet as well.
|
|
|
|
No longer used with the parser rewrite.
|
|
This completely removes any need to look behind for already parsed records,
and adds a second pass that assigns stats to records that did not have that
information known at the time of parsing.
With the testing data, this changed the number of records with no stat
information from approximately 1/6 to none (!).
The parser is also much cleaner now, at least as much as a parser built around
regular expressions is clean. Apparently the late night coding me really likes
nesting closures.
This closes #7.
This closes #13.
|
|
The PC chart is removed, as records being linked to specific characters could
never be released publically.
The records are being postprocessed, as information about stats for unknown
records could have been logged after the record (ie. STAT on LOGOUT)
The processed records are available for download in JSON format. Currently,
this is only accessible by calling parser.createBlobLink() from the console.
The process can be quite resource intensive.
The file should be saved with "scrubbed" appearing somewhere in the filename.
The loader now properly informs its each method of the current file and the
number of files.
Files with "scrubbed" in the filename are interpreted as scrubbed logfiles and
will be loaded in as JSON directly.
|
|
|