Migrate to use isic-metadata 1.0.0

isic_cli/cli/metadata.py

@@ -5,10 +5,11 @@
 from pathlib import Path
 import sys
 
+
 import click
 from click.types import IntRange
 from humanize import intcomma
-from isic_metadata.metadata import MetadataBatch, MetadataRow
+from isic_metadata.metadata import MetadataBatch, MetadataRow, convert_errors
 from isic_metadata.utils import get_unstructured_columns
 from pydantic import ValidationError
 from rich.console import Console
@@ -30,15 +31,22 @@ def metadata(obj):
 @metadata.command(name="validate")
 @click.argument(
     "csv_file",
-    type=click.File("rb"),
+    type=click.File("r"),
 )
 def validate(csv_file: io.BufferedReader):
     """Validate metadata from a local csv."""
-    # These imports are slow, inline them.
-    import pandas as pd
-
     console = Console()
-    df = pd.read_csv(csv_file, header=0)
+
+    # get number of rows in csv
+    num_rows = sum(1 for _ in csv_file)
+    csv_file.seek(0)
+
+    reader = csv.DictReader(csv_file)
+    headers = reader.fieldnames
+
+    if not headers:
+        click.secho("No rows found in csv!", fg="red")
+        sys.exit(1)
 
     # batch problems apply to the overall csv and can't be computed without looking at the
     # entire csv.
@@ -47,20 +55,26 @@ def validate(csv_file: io.BufferedReader):
     # keyed by column, message
     column_problems: dict[tuple[str, str], list[int]] = defaultdict(list)
 
-    for i, (_, row) in track(enumerate(df.iterrows(), start=2), total=len(df)):
+    batch_items: list[MetadataRow] = []
+
+    # start enumerate at 2 to account for header row and 1-indexing
+    for i, row in track(
+        enumerate(reader, start=2), total=num_rows, description="Validating metadata"
+    ):
+        if row.get("patient_id") or row.get("lesion_id"):
+            batch_items.append(
+                MetadataRow(patient_id=row.get("patient_id"), lesion_id=row.get("lesion_id"))
+            )
         try:
-            MetadataRow.model_validate(row.to_dict())
+            MetadataRow.model_validate(row)
         except ValidationError as e:
-            for error in e.errors():
+            for error in convert_errors(e):
                 column = error["loc"][0]
                 column_problems[(column, error["msg"])].append(i)
 
     try:
         MetadataBatch(
-            items=[
-                MetadataRow(patient_id=row.get("patient_id"), lesion_id=row.get("lesion_id"))
-                for _, row in df.iterrows()
-            ]
+            items=batch_items,
         )
     except ValidationError as e:
         for error in e.errors():
@@ -105,7 +119,7 @@ def validate(csv_file: io.BufferedReader):
     else:
         click.secho("No structural errors found!", fg="green")
 
-    unstructured_columns = get_unstructured_columns(df)
+    unstructured_columns = get_unstructured_columns(headers)
     if unstructured_columns:
         table = Table(title="Unrecognized Fields")
         table.add_column("Field", justify="left", style="cyan", no_wrap=True)

setup.py

@@ -42,11 +42,9 @@
         # We expect girder-cli-oauth-client to drop oob support in the future
         "girder-cli-oauth-client<1.0.0",
         "humanize",
-        "isic-metadata>=0.4.0",
+        "isic-metadata>=1.0.0",
         "more-itertools",
-        "numpy",
         "packaging",
-        "pandas",
         "requests",
         "retryable-requests",
         "rich",

tests/test_cli_metadata.py

@@ -32,8 +32,8 @@ def test_metadata_validate(runner, cli_run):
         result = cli_run(["metadata", "validate", "foo.csv"])
 
     assert result.exit_code == 1, result.exception
-    assert re.search(r"Invalid diagnosis.*foo", result.output), result.output
-    assert re.search(r"Invalid sex.*bar", result.output), result.output
+    assert re.search(r"Unsupported value for diagnosis: 'foo'.", result.output), result.output
+    assert re.search(r"sex.*Input should be 'male' or 'female'", result.output), result.output
 
 
 def test_metadata_validate_lesions_patients(runner, cli_run):