User Manual

User Manual: Pdf

Open the PDF directly: View PDF PDF.
Page Count: 3

DownloadUser Manual
Open PDF In BrowserView PDF
Data
get_data​ is used to fill the database (sqlite) with bug data from defects4j and associated
gzoltar spectra.
To load data for a defects4j-bug (e.g. Time 1) into the db use ipython:
from getdata.get_data import GetData
d = GetData()
d.get_data(project=‘Time’, bug=1)
If no bug is given, it will get all bugs of the project, if no project is given as well, all bugs of all
projects are processed.

Model
model ​is used to build decision trees with Weka and to parse and analyze its result. It also
calculates the suspiciousness for each method.
To generate the model and get the methods ordered by suspiciousness use ipython:
from model import Model
m = Model(‘Time’, 1)
### takes a while, builds the decision tree and parses it
results, methods, mid_2_list_id = m.get_suspicious_methods()
=>​ ​results​ contains the ​suspiciousness​ values for each ​mid ​(method id) and ​methods
contains the ​method names​ (= Java-File name and line number of method) and their
lengths​ in the same order (by highest suspiciousness).
mid_2_list_id​ is a mapping from each mid to its place in the list ​methods.
Evaluation
In evaluateresult.py in __main__:
bugs = [ {'project': 'Time', 'bug': 1} ]
add those defects4j bugs you want to evaluate to the list and call:
print_results(bugs, 'test')
where ‘test’ is a file to which the output is written (it’s also printed on the console). If the
needed ​data​ is ​not​ found​ in the database​ (get_data.db) the data for that bug ​will be
generated​ (defects4j, gzoltar, weka, …).
OR in __main__:
evaluate_project('Mockito', 'test'')

to evaluate the ​whole​ ​project ​(here Mockito). With that command ​only​ those ​bugs​ of the
given project that are ​already in the database​ are evaluated.
Then call:
python evaluateresult.py
and you should get something like:
Results for bug Mockito 12
=================================
Buggy method: org/mockito/internal/util/reflection/GenericMaster.java, line 16
Rank: 1.0
Best rank: 1
Worst rank: 2
Wasted effort: 3
10 most suspicious methods ordered by suspiciousness:
0.568182, org/mockito/internal/util/reflection/GenericMaster.java, line 16
0.568182, org/mockito/internal/util/reflection/GenericMaster.java, line 8
0.535354, org/mockito/ArgumentCaptor.java, line 70
0.535354, org/mockito/ArgumentMatcher.java, line 57
0.535354, org/mockito/internal/matchers/CapturingMatcher.java, line 16
0.535354, org/mockito/internal/progress/HandyReturnValues.java, line 17
0.535354, org/mockito/ArgumentCaptor.java, line 122
0.535354, org/mockito/internal/configuration/DefaultAnnotationEngine.java, line 69
0.489899, org/mockito/internal/util/reflection/AccessibilityChanger.java, line 16
0.489899, org/mockito/internal/util/reflection/FieldSetter.java, line 19

Installation:
Requirements:
Linux with
- Oracle JDK 7
- https://bitbucket.org/rjust/fault-localization-data/overview​ (for gzoltar and
defects4j)
- Weka https://www.cs.waikato.ac.nz/ml/weka/

Either you download the program from moodle or clone the project from github:
git clone ​https://github.com/Florian1990/ml4swt
git checkout final
To install the required packages for python:
pip install -r requirements.txt

If you want to use a prefilled database you can use get_data.db from:
https://web.tresorit.com/l#H9wKJ5AWg18m9078-1FLKA​ (this URL is accessible until
20th of February)
Source Exif Data: 
File Type                       : PDF
File Type Extension             : pdf
MIME Type                       : application/pdf
PDF Version                     : 1.5
Linearized                      : Yes
Producer                        : Skia/PDF m65
Page Count                      : 3
EXIF Metadata provided by EXIF.tools

Navigation menu