filter: --query fails when the .str accessor is used on a column

[victorlin]

First reported in a discussion post.

Current Behavior

Given a metadata file with a string column column, any query using column.str results in an error.

cat >metadata.tsv <<~~
strain	column
SEQ_1	value1
SEQ_2	value2
SEQ_3	value3
~~

augur filter \
  --metadata metadata.tsv \
  --query "column.str.startswith('value')" \
  --output-strains filtered_strains.txt
# ERROR: Internal Pandas error when applying query:
# 	'NoneType' object has no attribute 'str'
# Ensure the syntax is valid per <https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#indexing-query>.

This is because extract_variables added in 2ead5b3 is unable to properly extract column as a variable name.

Expected behavior

The filter should run successfully.

augur filter \
  --metadata metadata.tsv \
  --query "column.str.startswith('value')" \
  --output-strains filtered_strains.txt
# 0 strains were dropped during filtering
# 4 strains passed all filters

Possible solutions

1. Fix extract_variables to support the column.str accessor. This isn't trivial since the error message doesn't provide the column name, which is what the function relies on for extracting variables from the query string.
2. Remove extract_variables and attempt type conversion of all metadata columns to numeric.

After realizing that extract_variables isn't easy to implement properly, I'm leaning towards option 2.

Affected versions

This bug impacts Augur version 22.2.0.