search

sansadjadee / fighting-layout-bugs

Automatically exported from code.google.com/p/fighting-layout-bugs
Apache License 2.0
0 stars 0 forks source link

Suggestions for 2 new Detectors #6

Open GoogleCodeExporter opened 9 years ago

GoogleCodeExporter commented 9 years ago 
The site http://arngren.net is a rich environment for finding layout bugs. The 
site includes various bugs with the current FLB code discovers. It also has 
some bugs FLB doesn't yet seem to detect. 

Here are the suggestions for new detectors:

TextOverlapsText

TextInadvertentlyWrapping

I've attached a document with some screenshots that highlight the problems for 
the arngren.net web site. 

Here are some initial thoughts on ways to detect both problems, these ideas are 
half-baked and would need to be refined :)

How to detect TextOverlapsText:
- Identify each segment of text (I assume each would be contained in a distinct 
WebElement).
- For each segment, calculate the x,y offset of the 4 points of the rectangle 
that contains the text (top left, top right, bottom left, bottom right)
- For each rectangle determine if it overlaps another text rectangle.

I'm limiting the formula to text that is horizontal or vertical.

How to detect TextInadvertentlyWrapping:
I've several ideas related to this:
1. length of the text on each line. If the last line of text has only a few 
characters e.g. 1 or 2, and there is no preceding new line character - flag as 
a possible 'InadvertentWrap'. This may generate various false positives, so 
would need to be checked against a body of sample real-world web content.
2. dimensions of the text area as text size is incremented and decremented. For 
single words or short phrases, they generally are 1 line high. The relative 
height of each text area should remain roughly constant as the text size is 
increased or decreased. When the height dimension changes disproportionately 
perhaps the number of lines needed to display the text has changed. Sometimes 
this is the correct behaviour and to be expected. Other times, for instance 
with the http://arngren.net web site, the height dimension sometimes decreases 
relatively (compared to the rest of the text elements) while the overall 
dimension increases. I suspect we'd need to apply a combined algorithm that 
uses idea 1 (number of characters on last line) with varying the text size.

I'm sure these suggested algorithms have flaws in them and wouldn't be 
sufficiently accurate to provide a good signal to noise ratio (noise coming 
from false positives, etc). However they might stimulate ideas so you can 
devise better algorithms :)

PS: I cannot submit this 'issue' as a feature enhancement in the UI. So please 
change it to a feature enhancement if you have permission to do so.

Julian Harty

-

Original issue reported on code.google.com by julianharty on 6 May 2012 at 10:10

Attachments:

GoogleCodeExporter commented 9 years ago

Original comment by michael.tamm2 on 13 May 2012 at 7:15