| |
| META TOPICPARENT |
name="DuckySherwood" |
Ducky's Homework / Research |
|
<
< | I am working for Gail Murphy, looking at programmer productivity. I've gone through several ideas for thesis topics so far: |
>
> | I am working for Gail Murphy, looking at programmer productivity. : |
| |
-
discovering interesting patterns in programmer/IDE interaction by data mining interaction logs (got stumped trying to find things without knowing what I was looking for; also most data mining techniques involve a small amount of data for a large number of subjects, not a large amount of data for a small number of subjects)
-
running a user study of novice pairs and experienced pairs of programmers, seeing how long they took to complete a set of ~4 tasks, and figuring out what the successful programmers did that the unsuccessful programmers didn't (realized that randomness plays such a large factor that I wasn't going to find anything interesting)
-
implementing tabbing behaviour in Eclipse similar to tabbing behaviour in Firefox, giving subjects a little lecture on doing breadth-first instead of depth-first search, and timing new-tab vs. old-tab behavior (someone else already implemented "Firefox" tabbing behaviour)
|
|
<
< |
- implementing a visualization of coders'navigation history
|
>
> |
-
implementing a visualization of coders'navigation history
- studying how tool support affects navigation patterns
|
| | |
| |
|
|
>
> | |
| | Current TBDs |
|
<
< |
-
blog on JML
- read a bunch of papers on navigation
|
| |
- try out IDEs to see how they do tabbing
- NetBeans: fwd/back buttons; moving to new file always opens new tab; tabs don't move, left/rt arrow to move view of tabs
- JCreator (Windows only)
|
|
<
< | |
>
> | |
| |
-
- Gel
 (Windows only, dormant)
|
|
>
> | |
| |
-
- BlueJ -- teaching tool; not sure if it scales
|
|
<
< |
- ask Mik/Rob re visualization
|
| |
- Rob's thesis mentions that 58-81% of web navigation is to previously seen pages, but passes on saying what % is for source. That would be interesting to try to pull out of Mylar data. (Did I finish his thesis?)
- Bannon et al: "[D]igressions are frequent. When a record is made of these commands as they occur temporarily, as in the history list, information on the tasks and goals of the users is lost."
- good stuff here: http://portal.acm.org/results.cfm?coll=GUIDE&dl=GUIDE&CFID=19182022&CFTOKEN=49296027
|
|
<
< | links:
daikon |
| |
Current hypotheses
- H0: Most of time "wasted" is due to false hypotheses. (is there a ko paper that says so?)
- H1: People make false hypotheses frequently.
|
|
<
< |
- H2: Breadth-first search when seeking verification for a hypothesis is on average faster than a depth-first search.
- H3: The "three hypotheses" approach is useful.
- H4: Better tools to see nav history are useful.
|
>
> |
- H2: Breadth-first navigation when seeking verification for a hypothesis is on average faster than a depth-first navigation
- H3: The following things will result in users using BFN more often.
- H3.1: tabs staying in once place
- H3.2: tabs being containers for a task rather than a file (more like Firefox)
- H3.3: making the back/forward button history only contain what was looked at in that tab, vs. everything in the global history
- H3.3: marking "special" places in the history drop-down with icons
- classes/methods reached via a search
- classes/methods reached via the package explorer
- classes/methods reached via
- (Question: what about reached from Outline View? Class Hierarchy view? QuickOutline)
- (Question: do I want to insert a right arrow if they did declaration and left if they did references? What about a refs/declr search?)
- H3.4: telling people that BFN is a better approach
- (H3.5: telling people to write down three hypotheses)
|
| | |
|
<
< |
- H10: code with JML is faster to understand than code without
|
| |
Research |
| |
follow-on to the camel |
|
>
> | Links to software tools
daikon |
| | |
Copyright © 2008-2025 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.