Bug? Expression Engine update Inconsistent Figures?

[paullee]

Hi, just notice a strange phenomenon...

Using expression to link objects - in example attached - object's placement e.g. Axes.Placement.Rotation.Axis.x linking the Section Plane to Axis. So far so good.

But when I click the property field with the expressions inside, the figures in these fields returned something slightly difference each time
So the section plane slightly shift, shake....

In the Axis-xyz fields, it originally read -0.6, or Axis field read -0.58,-0.58,-0.58.

Clicking them, it sometime return -0.57 for 1 field or another ... then the section plane shake!

Missing something?

phpBB [video]

Screenshot from 2018-04-21 16-45-42.png (165.92 KiB) Viewed 1189 times

Screenshot from 2018-04-21 16-46-14.png (173.94 KiB) Viewed 1189 times

Screenshot from 2018-04-21 16-48-02.png (206.92 KiB) Viewed 1189 times

0.17.13509 (Git) AppImage on Fedora 27
(OCC 7.1.0)

OS: Linux
Word size of OS: 64-bit
Word size of FreeCAD: 64-bit
Version: 0.17.13509 (Git)
Build type: None
Branch: master
Hash: 0258808ccb6ba3bd5ea9312f79cd023f1a8671b7
Python version: 2.7.6
Qt version: 4.8.6
Coin version: 4.0.0a
OCC version: 7.1.0
Locale: English/HongKong (en_HK)

[DeepSOIC]

Hi!
I have my number of decimals up by quite a bit than default, and I don't experience the issue. Seems to be rounding getting in the way somehow.

placement-expression-no-problem.png (105.25 KiB) Viewed 1176 times

OS: Windows 8
Word size of OS: 64-bit
Word size of FreeCAD: 64-bit
Version: 0.17.13509 (Git)
Build type: Release
Branch: releases/FreeCAD-0-17
Hash: 0258808ccb6ba3bd5ea9312f79cd023f1a8671b7
Python version: 2.7.14
Qt version: 4.8.7
Coin version: 4.0.0a
OCC version: 7.2.0
Locale: Russian/RussianFederation (ru_RU)

[paullee]

Thanks, it seem to be the crux of problem

Previously set the unit decimal place to 0.1mm (much more than enough for Architecture in fact), now set to some 12 places seem solve the problem.


But, this should be a Bug - the internal calculation should not be affected - I thought the decimal place setting is sort of 'presentation'
Maybe other problems I had run into related to this...


Anybody can look into this?

Thanks again.

Screenshot from 2018-04-22 00-09-56.png (186.75 KiB) Viewed 1159 times

Screenshot from 2018-04-22 00-11-20.png (168.62 KiB) Viewed 1159 times

[chrisb]

You should have changed the units instead of the precision. I haven't thought about this before, but as soon as angles come into play, it's not enough to have only 1 decimal.

[paullee]

You should have changed the units instead of the precision. I haven't thought about this before, but as soon as angles come into play, it's not enough to have only 1 decimal.

Thanks.

I have been using mm, with 1 decimal place, which is far more than enough for building or interior design works. Architecture / I.D. usually works with mm / cm (or inch or imperial system). Not quite handy, if use m with 3 decimal place, say 0.008m to represent thickness of plywood instead of 8mm

Now I set precision to say 12 decimal place, it also sound ridiculous but solve the problem to read 8.000000000000000mm plywood.


But yes then I know when that 0.1 precision come into play on other scenario, rotation e.g., which is sometime radian, it means 1 or 2 degree.


I think this is a problem / bug - as I said, I thought the 'precision' setting is a representation only.

[paullee]

BTW, I remember I still get reading of a vertex of e.g. ArchWall / Underlying Sketch as something 98.33223525235235215135mm - even I think I had long time no changing (not realizing the 'importance') of the decimal place from that 1 decimal place setting.

So it seem the setting is not consistently having 'effect' on different places?

[DeepSOIC]

A better implementation could be to set number of meaningful digits, rather than number of digits after decimal point. Then it would say "8.0000 mm" for thickness and "10000 mm" for length without any problem. It has already been reported, that setting Clone's scale to 1:1000 is impossible with standard unit preferences.

In the end, even though it should be only a display thing, nailing it in every possible number editing field can be a big coding effort. So my preferred solution (as user) is still to set a fair bit of extra precision, just to be safe.

[ulrich1a]

I think this is a problem / bug - as I said, I thought the 'precision' setting is a representation only.

I consider it also a bug. In some places it is not possible to set accurate numbers from the Gui. Consider the case, that you want to have the length of a cube pi mm = 3.141592653589793 mm.
FreeCAD knows pi, so one can type in PartDesign pi mm for the length in the Gui-dialog for an additive cube. You end with a length set to 3.14 mm, because the Gui dialog cuts independent of a setting the calculated length after two digits after the decimal point.

Trying to set it in the property editor yields 3,141590000000 mm.
Using python to set it, yields 3,141592653590 mm. This stays until the property is touched, than is it again 3,141590000000 mm.
Measuring the part delivers 3.141592741013 mm. This seems to be caused by the internal accuracy of the cad-kernel and is expected.
After touching the number its 3.14159011... mm.
A problem is touching the property in the Gui, which feeds back the rounding to the part.

Ulrich

[paullee]

A better implementation could be to set number of meaningful digits, rather than number of digits after decimal point. Then it would say "8.0000 mm" for thickness and "10000 mm" for length without any problem. It has already been reported, that setting Clone's scale to 1:1000 is impossible with standard unit preferences.

Thought for a while …

For use related to geographical matter, GeoData, Civil EngIneering, Arch Site related where one might need to have actual world coordinates (hugh difference in US or in Europe), but still precision of model may still be in mm or +-10mm … not sure how it could works.

BTW, currently setting to 12 decimal place have impact on performance?

[paullee]

National Geograhic Society indicate circumference of the earth is 40,075 km

i,e. 40,075,000,000 mm?

11 digits