TypeError: 'int' object is not a mapping in tsinfer infer commandline

[haneenih7]

Hello,

I am running the command infer from tsinfer, and I get the following error:

File "/ibex/sw/csi/tsinfer/0.2.2/el7.9_conda/miniconda3/bin/tsinfer", line 11, in
sys.exit(main())
File "/ibex/sw/csi/tsinfer/0.2.2/el7.9_conda/miniconda3/lib/python3.9/site-packages/tsinfer/main.py", line 5, in main
cli.tsinfer_main()
File "/ibex/sw/csi/tsinfer/0.2.2/el7.9_conda/miniconda3/lib/python3.9/site-packages/tsinfer/cli.py", line 499, in tsinfer_main
args.runner(args)
File "/ibex/sw/csi/tsinfer/0.2.2/el7.9_conda/miniconda3/lib/python3.9/site-packages/tsinfer/cli.py", line 143, in run_infer
ts = tsinfer.infer(
File "/ibex/sw/csi/tsinfer/0.2.2/el7.9_conda/miniconda3/lib/python3.9/site-packages/tsinfer/inference.py", line 291, in infer
ancestors_ts = match_ancestors(
File "/ibex/sw/csi/tsinfer/0.2.2/el7.9_conda/miniconda3/lib/python3.9/site-packages/tsinfer/inference.py", line 449, in match_ancestors
return matcher.match_ancestors()
File "/ibex/sw/csi/tsinfer/0.2.2/el7.9_conda/miniconda3/lib/python3.9/site-packages/tsinfer/inference.py", line 1401, in match_ancestors
ts = self.store_output()
File "/ibex/sw/csi/tsinfer/0.2.2/el7.9_conda/miniconda3/lib/python3.9/site-packages/tsinfer/inference.py", line 1477, in store_output
ts = self.get_ancestors_tree_sequence()
File "/ibex/sw/csi/tsinfer/0.2.2/el7.9_conda/miniconda3/lib/python3.9/site-packages/tsinfer/inference.py", line 1448, in get_ancestors_tree_sequence
self.convert_inference_mutations(tables)
File "/ibex/sw/csi/tsinfer/0.2.2/el7.9_conda/miniconda3/lib/python3.9/site-packages/tsinfer/inference.py", line 1232, in convert_inference_mutations
metadata = _update_site_metadata(site.metadata, constants.INFERENCE_FULL)
File "/ibex/sw/csi/tsinfer/0.2.2/el7.9_conda/miniconda3/lib/python3.9/site-packages/tsinfer/inference.py", line 150, in _update_site_metadata
return {"inference_type": inference_type, **current_metadata}
TypeError: 'int' object is not a mapping

Commandline:
tsinfer infer test.sample --num-threads 40

I had the same error when doing the inference in Jupyter using the tutorial codes and example.

[hyanwong]

Thanks for the bug report @haneenih7 - I'll look into this ASAP. I think it's something to do with the site metadata. Do you know what version of tskit you are using (import tskit; print(tskit.__version__))

[hyanwong]

Hi @haneenih7: can I clarity that in a new python3 session, the following tutorial example fails with the same error:

import tsinfer

with tsinfer.SampleData(sequence_length=6) as sample_data:
    sample_data.add_site(0, [0, 1, 0, 0, 0], ["A", "T"])
    sample_data.add_site(1, [0, 0, 0, 1, 1], ["G", "C"])
    sample_data.add_site(2, [0, 1, 1, 0, 0], ["C", "A"])
    sample_data.add_site(3, [0, 1, 1, 0, 0], ["G", "C"])
    sample_data.add_site(4, [0, 0, 0, 1, 1], ["A", "C"])
    sample_data.add_site(5, [0, 1, 2, 0, 0], ["T", "G", "C"])


inferred_ts = tsinfer.infer(sample_data)

[jeromekelleher]

I think this is the ongoing Zarr problem @hyanwong - are you on the latest release of tsinfer? Where are things with that issue @benjeffery?

[hyanwong]

Ah, thanks @jeromekelleher. I am on the latest tsinfer release, but I probably have an older version of zarr, as it works for me:

>>> inferred_ts = tsinfer.infer(sample_data)
>>> import zarr
>>> zarr.__version__
'2.8.3'

[hyanwong]

For reference, here is the zarr issue:#628

[benjeffery]

Where are things with that issue @benjeffery?

Next release of zarr will have the indexing fix zarr-developers/zarr-python#974 (comment)

For fill_value I think we need to specify as it is a larger issue. Will do a PR to that effect here.

[haneenih7]

@hyanwong @benjeffery @jeromekelleher Thank you all for the input

I am not using tskit. I am using tsinfer version 0.2.2

@hyanwong with the tutorial example you provided, it is giving the same issue.

[hyanwong]

@hyanwong @benjeffery @jeromekelleher Thank you all for the input

I am not using tskit. I am using tsinfer version 0.2.2

Installing tsinfer will have installed tskit under the hood: tskit is a required library for tsinfer.

@hyanwong with the tutorial example you provided, it is giving the same issue.

Thanks. I think this is the zarr issue (which also will have been installed by tsinfer, I suspect). In python, what do you get with

import zarr
zarr.__version__

You should be able to downgrade zarr from the command line (not within python) by

python3 -m pip uninstall zarr
python3 -m pip install zarr==2.11.0

Perhaps try that and see if tsinfer then works?

[jeromekelleher]

That's the only version of zarr you don't want, try python3 -m pip install "zarr<2.11"

[hyanwong]

Oops. Sorry @jeromekelleher - my bad!

[haneenih7]

@jeromekelleher @hyanwong Thank you very much guys for your fast response and support. I highly appreciate it.
I installed zarr<2.11 python3 -m pip install zarr==2.10.0
and now the provided example as well as the data example from the tutorial works 👍