MAINT: Remove empty features #171

This is gonna be kind of inconvient to apply to the feature
ranks, also. Think I'm going to go back and try an alternative approach
to all of this stuff (#172, #171, #58) on another branch.

qurro/_df_utils.py

@@ -145,24 +145,34 @@ def biom_table_to_sparse_df(table, min_row_ct=2, min_col_ct=1):
     return table_sdf
 
 
-def remove_empty_samples(biom_table):
-    """Removes samples with 0 counts for every feature from a BIOM table.
+def remove_empty(biom_table):
+    """Removes "empty" samples and features from a BIOM table.
 
        This will raise a ValueError if, after removing empty samples, the
        table's columns are empty (this will happen if all of the samples in
        the table are empty).
     """
-    logging.debug("Attempting to remove empty samples.")
-    filtered_table = biom_table.remove_empty(axis="sample", inplace=False)
+    # TODO make this also somehow alter the ranks so that these are removed
+    # from there? (so matching doesn't fail?)
+    # And while we're at it we should maybe also do the same thing for sample
+    # metadata tbh.
+    logging.debug("Attempting to remove empty samples and features.")
+    filtered_table = biom_table.remove_empty(inplace=False)
+
+    if filtered_table.shape[0] < 1:
+        raise ValueError("All of the current features in the table are empty.")
 
     if filtered_table.shape[1] < 1:
         raise ValueError("Found all empty samples with current features.")
 
-    sample_diff = filtered_table.shape[1] - biom_table.shape[1]
-    if sample_diff > 0:
-        logging.debug("Removed {} empty sample(s).".format(sample_diff))
-    else:
-        logging.debug("Couldn't find any empty samples to remove.")
+    for name, idx in (("feature", 0), ("sample", 1)):
+        diff = filtered_table.shape[idx] - biom_table.shape[idx]
+        if diff > 0:
+            logging.debug("Removed {} empty {}(s).".format(name, diff))
+        else:
+            logging.debug(
+                "Couldn't find any empty {}s to remove.".format(name)
+            )
 
     return filtered_table
 

qurro/generate.py

@@ -26,7 +26,7 @@
     replace_nan,
     validate_df,
     biom_table_to_sparse_df,
-    remove_empty_samples,
+    remove_empty,
     match_table_and_data,
     merge_feature_metadata,
 )
@@ -75,8 +75,8 @@ def process_input(
           using the provided extreme_feature_count. (If it's None, then nothing
           will be done.)
 
-       4. Calls remove_empty_samples() to filter samples without any counts for
-          any features from the BIOM table. This is purposefully done *after*
+       4. Calls remove_empty() to filter empty samples and features from the
+          BIOM table. This is purposefully done *after*
           filter_unextreme_features() is called.
 
        5. Converts the BIOM table to a SparseDataFrame by calling
@@ -145,7 +145,7 @@ def process_input(
     )
 
     # 4. Filter (now-)empty samples from the BIOM table.
-    filtered_biom_table = remove_empty_samples(feature_filtered_biom_table)
+    filtered_biom_table = remove_empty(feature_filtered_biom_table)
 
     # 5. Convert the BIOM table to a SparseDataFrame.
     unmatched_table = biom_table_to_sparse_df(filtered_biom_table)

qurro/tests/test_df_utils.py

@@ -9,7 +9,7 @@
     ensure_df_headers_unique,
     validate_df,
     replace_nan,
-    remove_empty_samples,
+    remove_empty,
 )
 
 
@@ -241,7 +241,7 @@ def test_replace_nan():
 
 
 def get_test_data():
-    """Returns test data for the remove_empty_samples() tests."""
+    """Returns test data for the remove_empty() tests."""
     feature_ids = ["F1", "F2", "F3", "F4", "F5", "F6", "F7", "F8"]
     sample_ids = ["Sample1", "Sample2", "Sample3", "Sample4"]
     table_data = np.array(
@@ -255,24 +255,24 @@ def get_test_data():
     return table_data, feature_ids, sample_ids
 
 
-def test_remove_empty_samples_basic():
-    """Tests remove_empty_samples() in the simple cases of removing 0, 1, and 2
+def test_remove_empty_basic():
+    """Tests remove_empty() in the simple cases of removing 0, 1, and 2
        empty sample(s).
     """
 
     # TRY REMOVING 0 SAMPLES
     data, fids, sids = get_test_data()
     table = biom.Table(data, fids, sids)
     # Check that, when none of the samples are empty, nothing is changed.
-    ftable = remove_empty_samples(table).to_dataframe()
+    ftable = remove_empty(table).to_dataframe()
     assert_frame_equal(ftable, table.to_dataframe())
 
     # TRY REMOVING 1 SAMPLE
     # Zero out Sample3 (it only has one count, for F1)
     data[0][2] = 0
     table = biom.Table(data, fids, sids)
     # Check that just the one empty sample (Sample3) was removed.
-    ftable = remove_empty_samples(table).to_dataframe()
+    ftable = remove_empty(table).to_dataframe()
     assert_array_equal(ftable["Sample1"], data[:, 0])
     assert_array_equal(ftable["Sample2"], data[:, 1])
     assert_array_equal(ftable["Sample4"], data[:, 3])
@@ -284,7 +284,7 @@ def test_remove_empty_samples_basic():
     # Now, zero out Sample4 (it only has one count in F4)
     data[3][3] = 0
     table = biom.Table(data, fids, sids)
-    ftable = remove_empty_samples(table).to_dataframe()
+    ftable = remove_empty(table).to_dataframe()
     assert_array_equal(ftable["Sample1"], data[:, 0])
     assert_array_equal(ftable["Sample2"], data[:, 1])
     assert "Sample3" not in ftable.columns
@@ -293,10 +293,10 @@ def test_remove_empty_samples_basic():
     assert len(ftable.index) == len(fids) == 8
 
 
-def test_remove_empty_samples_allempty():
-    """Tests remove_empty_samples() when all samples in the table are empty."""
+def test_remove_empty_allempty():
+    """Tests remove_empty() when all samples in the table are empty."""
 
     _, feature_ids, sample_ids = get_test_data()
     table = biom.Table(np.zeros(32).reshape(8, 4), feature_ids, sample_ids)
     with pytest.raises(ValueError):
-        remove_empty_samples(table)
+        remove_empty(table)