forked from Human-AI-Interaction-UBC/EMDAT
-
Notifications
You must be signed in to change notification settings - Fork 1
/
bug_fix.txt
47 lines (37 loc) · 5.12 KB
/
bug_fix.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
Record of Bugs Found so far during the Second and AOI Phases of the EMDAT_testing Project
Date of Detection Date of Resolution Pertinent Part of Code Description
1. 04 May 2017 05 May 2017 vector_difference in In the if-else statement block, the angle for vertical eye movement
EMDAT_core\geometry.py and that for horizntal eye movement were placed under the wrong condition,
lines: 39-42 swapping the values.
2. 08 May 2017 10 May 2017 second value assignment to The attribute was actually reassigned the value, len(distances), in that line,
the numdistances attibute in overwriting the previously assigned value, len(valid_distance_data), in line
__init()__ of Segmemt.py 188 in the same __init()__. The former value concerns fixation, while the latter,
at line 225 head distance related to pupil data. The reassigned attribute value was passed to a
Scene object, where it was used for computation of the mean head distance in line 447
of Scene.py, affecting the computed mean head distance. A diffent variable was created
for demarcation.
3. 10 May 2017 11 May 2017 calc_rel_angles in Segment.py Due to the sign convention and range of vector_difference, which is defined in
geometry.py and was invoked in calc_rel_angles (prior to the fix), computed realtive
differences were not always internal angles. calc_rel_angles was redifined.
4. 11 May 2017 11 May 2017 calc_distances, Not strictly a bug in that it had no adverse effect on the function of application.
calc_abs_angles, and However, those methods were never used anywhere in the code, so that they were
calc_rel_angles in Scene.py removed.
5. 24 May 2017 24 May 2017 calc_rel_angles in Segment.py At least for one scene, apparently due to float accuracy, a dot product of normalized
last and next vectors exceeded the lower bound of -1, inducing error in the subsequent
computation of arccosine that uses the value. The fix was to use if statements for cases
where a dot product vlaue in float falls outside of the domain of arc cosine.
6. 25 May 2017 26 May 2017 computation coede for saccade Due to an apparent syntax error (saccade_data != None, which is still true when
features in the constructor saccade_data = []) in an if-statement, an empty list was passed to build-in max
of Segment.py min funcitons, resulting in an error. An additonal clauee len(saccade_data) > 0 was
implemented. Another change required for the fix is value assignment to
the numsaccade attribute in the else clause, which was originally missing.
7. 25 May 2017 29 May 2017 calc_rel_angles in Segment.py The vectors passed to arc cos in the method are first normalized by dividing with the square
root of the dot product of each vecor with itself, so that a zero vecor resulted in a division
by zero. It is stipulated that if one or both vectors are zero, the relative angle is zero.
8. 2 June 2017 7 June 2017 merge_aoistats in Scene.py There were extraneous entries in the feature name row and _allsc rows in the exproted features file,
causing misalignment of names and their values. The cause was misuse of the built-in lstrip in the
function in question. An alternative method of taking a substring was applied.
Notes:
1) Date of Resolution refers to the time when the effectiveness of applied fix was
,if possible, confimred by EMDAT_teting available at https://github.com/dtoker/EMDAT_testing;
2) The line numbers mentioned refer to those at the time of detection of the bug.