yt code_length unit inconsistent with Ramses unit_l

[fredt00]

Bug report

Bug summary
I experience the same issue as #2866: the unit length conversion for a non cosmological simulation is off by a factor of the boxlength.

Code for reproduction

print(ds.arr(1,"code_density").in_cgs().value/ds.parameters['unit_d'])
print(ds.arr(1,"code_length").in_cgs().value/ds.parameters['unit_l'])
print(ds.parameters['boxlen'])

Actual outcome

The length conversion is off by a factor of 150, the value of the box length. However, the density conversion seems to be correct...

1.0
150.0
150.0

I was wondering if anyone had a simple work around for this? I don't think I can just multiply any results with length units by factors of the box length since the density conversion is correct and includes length**3.

Version Information

• Operating System: macOS Sonoma 14.1.1
• Python Version: 3.11.6
• yt version: 4.3.0 (installed via pip)

[welcome]

Hi, and welcome to yt! Thanks for opening your first issue. We have an issue template that helps us to gather relevant information to help diagnosing and fixing the issue.

[matthewturk]

Thanks for bringing this up -- I am sorry we didn't address it before! Perhaps @cphyc can provide some insight?

[cphyc]

I'm happy to help but I'm 200% busy until February 9 and won't have time until then. Most likely, these two lines are the right entry points:

yt/yt/frontends/ramses/data_structures.py

Line 977 in 1e90fd7

setdefaultattr(self, "length_unit", self.quan(length_unit * boxlen, "cm"))

yt/yt/frontends/ramses/data_structures.py

Line 1046 in 1e90fd7

self.domain_right_edge = np.ones(3, dtype="float64")

The first sets the length_unit, so 1 * length_unit equals the box size. The second sets the size of the box.
I wonder if the fix is simply to replace the latter by

        self.domain_right_edge = np.full(3, self.parameters["boxlen"], dtype="float64")

Could you try and let us know if this works? If it does, we'd be happy for you to submit a PR, or I can do it in a couple of weeks.

[fredt00]

Hi,

Thanks for the quick reply. I tried changing the line you suggested, and this does mean the correct domain_right_edge is returned when loading the data:

yt : [INFO     ] 2024-01-24 16:11:33,248 Parameters: domain_right_edge         = [150. 150. 150.]

However, the ratio code_length/unit_l is still incorrect by a factor of boxlen and I get the same offset in my plots.

I also tried removing the factor of boxlen in line 977 to give:

setdefaultattr(self, "length_unit", self.quan(length_unit , "cm")) 

I'm completely new to both Ramses and yt so I'm not sure if it was logical to do this, but this seems to completely break any plots I make of the gas.