(0026280)
Robert Maynard (developer)
2011-04-28 15:16

It looks like it actually sets all negative values to 0.0 if you are asking for log scale.

So in the usecase that you had negative, positive and zero values you would want all zero values to stay 0.0 and the negative values to become two orders of magnitude smaller than the smallest non zero positive value?

(0026281)
Greg Weirs (developer)
2011-04-28 15:36

log fails for any non-positive number, including zero, so something is being done to handle 0.

What I have observed is a surface where the values are positive, and a flat region where some small value has replaced negative numbers, and at the border between there are spikes below the flat region. I interpret the spikes as small positive values that are less than the replacement value for negative values. Small positive numbers are valid and not unusual, this is not a rare, academic case.

(0026282)
Robert Maynard (developer)
2011-04-28 15:44

I double checked all values less than or equal to zero are set to zero. Negative values are not given a special positive value currently.

(0026283)
Greg Weirs (developer)
2011-04-28 16:13

I don't understand why you don't get an error when you take the log of 0, but the desired end result is that negative and zero values should be reset to two orders of magnitude below the smallest positive value (before taking the log). Then take the log of the whole field.

(0028061)
Robert Maynard (developer)
2012-02-06 15:16

we don't take the log of 0, we just set the value to 0 instead of doing the log calculation.

This request functionality is doable but will slow down reading as we will have to iterate all the values twice.

(0028062)
Greg Weirs (developer)
2012-02-06 16:00

The problem is that in log space, the value 0 is often in the middle of the data. Setting to a value smaller than the smallest positive value ensures that the reset data is a minimum value in log space.

In a previous comment I said this resetting shouldn't be done for density and temperature. It turns out that 0 is often a value in these two fields, and when log scaling is selected, these should also be reset to two orders of magnitude below the minimum positive value.

The reduced reading speed is acceptable (since it only happens once.)

(0028063)
Robert Maynard (developer)
2012-02-06 16:04

So to clarify all properties must now have zero and negative values set to two orders of magnitude less than the smallest positive value. Than take the log of the whole field.

(0037965)
Kitware Robot (administrator)
2016-08-12 09:58

Resolving issue as `moved`.

This issue tracker is no longer used. Further discussion of this issue may take place in the current ParaView Issues page linked in the banner at the top of this page.

Issue History
Date Modified	Username	Field	Change
2011-04-27 12:41	Greg Weirs	New Issue
2011-04-28 13:44	Robert Maynard	Assigned To	=> Robert Maynard
2011-04-28 13:44	Robert Maynard	Status	backlog => tabled
2011-04-28 15:16	Robert Maynard	Note Added: 0026280
2011-04-28 15:36	Greg Weirs	Note Added: 0026281
2011-04-28 15:44	Robert Maynard	Note Added: 0026282
2011-04-28 16:13	Greg Weirs	Note Added: 0026283
2011-06-16 13:10	Zack Galbreath	Category	Feature Request => Feature
2012-02-06 14:58	Alan Scott	Status	backlog => todo
2012-02-06 15:11	Alan Scott	Project	=> Sandia
2012-02-06 15:11	Alan Scott	Type	=> incorrect functionality
2012-02-06 15:16	Robert Maynard	Note Added: 0028061
2012-02-06 16:00	Greg Weirs	Note Added: 0028062
2012-02-06 16:04	Robert Maynard	Note Added: 0028063
2012-05-31 10:23	Utkarsh Ayachit	Status	todo => backlog
2016-08-12 09:58	Kitware Robot	Note Added: 0037965
2016-08-12 09:58	Kitware Robot	Status	backlog => closed
2016-08-12 09:58	Kitware Robot	Resolution	open => moved

Notes
(0026280) Robert Maynard (developer) 2011-04-28 15:16	It looks like it actually sets all negative values to 0.0 if you are asking for log scale. So in the usecase that you had negative, positive and zero values you would want all zero values to stay 0.0 and the negative values to become two orders of magnitude smaller than the smallest non zero positive value?

(0026281) Greg Weirs (developer) 2011-04-28 15:36	log fails for any non-positive number, including zero, so something is being done to handle 0. What I have observed is a surface where the values are positive, and a flat region where some small value has replaced negative numbers, and at the border between there are spikes below the flat region. I interpret the spikes as small positive values that are less than the replacement value for negative values. Small positive numbers are valid and not unusual, this is not a rare, academic case.

(0026282) Robert Maynard (developer) 2011-04-28 15:44	I double checked all values less than or equal to zero are set to zero. Negative values are not given a special positive value currently.

(0026283) Greg Weirs (developer) 2011-04-28 16:13	I don't understand why you don't get an error when you take the log of 0, but the desired end result is that negative and zero values should be reset to two orders of magnitude below the smallest positive value (before taking the log). Then take the log of the whole field.

(0028061) Robert Maynard (developer) 2012-02-06 15:16	we don't take the log of 0, we just set the value to 0 instead of doing the log calculation. This request functionality is doable but will slow down reading as we will have to iterate all the values twice.

(0028062) Greg Weirs (developer) 2012-02-06 16:00	The problem is that in log space, the value 0 is often in the middle of the data. Setting to a value smaller than the smallest positive value ensures that the reset data is a minimum value in log space. In a previous comment I said this resetting shouldn't be done for density and temperature. It turns out that 0 is often a value in these two fields, and when log scaling is selected, these should also be reset to two orders of magnitude below the minimum positive value. The reduced reading speed is acceptable (since it only happens once.)

(0028063) Robert Maynard (developer) 2012-02-06 16:04	So to clarify all properties must now have zero and negative values set to two orders of magnitude less than the smallest positive value. Than take the log of the whole field.

(0037965) Kitware Robot (administrator) 2016-08-12 09:58	Resolving issue as `moved`. This issue tracker is no longer used. Further discussion of this issue may take place in the current ParaView Issues page linked in the banner at the top of this page.

Relationships