Conditional sampling doesn't work if the model has a CustomConstraint

[npatki]

Environment Details

• SDV version: 0.13.1
• Python version: 3.7

Error Description

Any time you create a tabular model using a CustomConstraint, you cannot sample it with conditions. (Sampling entire rows -- without conditions -- works.)

Currently, we only using reject sampling, so it should always be possible to sample an entire row and then filter out invalid conditions.

Steps to reproduce

Follow the code in issue #560 .

Current Workaround

It's possible to sample entire rows without constraints. So the workaround is to call sample(), and then manually delete any invalid rows that are returned. Note: It may take more tries to get the desired output, so you may want to try sampling a greater number of rows that you want.

Other notes

• Until we figure out how to fix this, it would at least be nice to give a better error message to users (filed in Detailed error message when conditional sampling on a model with CustomConstraint #694 )
• Related user issues: How to handle conditional sampling when using CustomConstraint? #560 , [KeyError] Conditional sampling with CustomConstraint  #687