Accessing TimeAveragingModifier output of LAMMPS compute from ComputePropertyModifier in Python

Dear Ovito devs,

I am right now trying to make a Python script to concatenate averaging compute outputs from LAMMPS from a dump file and the computation of properties to save some time.

Say I have a compute named ABC that outputs a per-atom vector in a dump file. The dump file columns will then be titled c_ABC[1] c_ABC[2] etc. When using the TimeAveragingModifier modifier, I can compute the per-atom averages by using

TimAveMod = TimeAveragingModifier(
    operate_on=["property:particles/c_abc[1]", "property:particles/c_abc[2]"]

Now say I want to operate on the output of TimAveMod and compute properties. The manual states that the fields will have Average appended to their name. Yet when doing:

ComProMod= ComputePropertyModifier(

Which is the name I get when exporting the file in the lammps dump format, but I get a RuntimeError from Data pipeline failure, as the Modifier does find the token c_abc1average. I tried several different writing like c_abc1average or c_abc1_average but none worked. Which is the correct one? The property is listed in the particles as 'c_abc[1] Average' but it is obviously not the name internally used.

Which is then the correct token to use for the parser to find it?

Hi Germain,

If the time-averaged particle property has the real name c_abc[1] Average, like you say, then the corresponding mangled property name to be used in the context of a Compute Property modifier is c_abc_1_Average.

The muParser component internally used by this modifier only supports variable identifiers made of alphanumeric characters and underscores, unfortunately. That’s why the modifier replaces all other characters in a property name with underscores to derive a valid mangled name, which can be referenced in the math expression.

Pro tip: You can use the GUI of OVITO to find out what the mangled name for a property is. First, insert a Compute Property modifier that creates a new dummy property with the desired name, e.g.:

Then insert another Compute Property modifier into the pipeline and open the list of available expression variables. Here it shows the mangled name:

1 Like

Yes makes sense. I noticed the replacement of characters with underscores but was unaware of the proper set of authorized characters since I didn’t know if the parser was in-build or not… The replacement of [ ] makes sense now.

The script works like a breeze now.

Obviously, the documentation was insufficient in this area. I have therefore expanded the Python docs of the ComputePropertyModifier a bit: ovito.modifiers — OVITO Python Reference 3.8.1 documentation

1 Like