| Age | Commit message (Collapse) | Author |
|---|
|
|
|
|
|
|
|
testing
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
according to my yet unwritten coding conventions:
1. no single line blocks
2. no empty lines without a good reason
|
|
verbose logging
|
|
Previously:
prefix #a does #b => ((a, b), [(a, b, "does (prefix)")])
now:
=> ((a, b), [(a, b, "prefix #a does #b")])
|
|
|
|
|
|
deleted
|
|
|
|
|
|
|
|
|
|
Auto suggest doesn't work for nodes loaded from local storage
Fixes #60
|
|
Works around a problem where you load rhizi with a small window size and then
enlarge it, since we don't dynamically change the svg size (and it is
fixed instead of filling a parent div or such) we lose access to a lot
of the screen (i.e. can't drag nodes there, can't see anything there,
etc).
|
|
1. move the sugg variable to textanalysis.js instead of the ui part
2. use a callback for the autocomplete widget (it supports three
options: array, string as url for a json returning service, or callback)
3. use a dictionary instead of an array to keep the list of suggestions.
We iterate on the dictionary instead of using a Trie or some other
datastructure in the callback. But the dictionary avoids iteration when
adding a word. Overall a loss in performance, but not noticeable.
Googling / memory suggests a Trie is what we want here, but I'm
postponing that for later. See:
http://ejohn.org/blog/javascript-trie-performance-analysis
|
|
|
|
|
|
|
|
|
|
|
|
Fixes #58
|
|
|
|
|
|
|
|
and reset early
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
request (GET?), which reloaded the page
|
|
|
|
|
|
no atomicity - nodes/links recorded as separate additions
We record both temp (state == 'temp') and not temp nodes,
can be filtered during analysis.
|
|
|
|
|
