docs: add streamlit app for showing the backend operation matrix

Author: cpcloud

docs/backend_info_app.py

@@ -0,0 +1,148 @@
+import csv
+from enum import Enum
+from pathlib import Path
+
+import pandas as pd
+import requests
+import streamlit as st
+
+st.set_page_config(layout='wide')
+
+
+@st.cache
+def load_ops_data():
+    response = requests.get(
+        'https://ibis-project.org/docs/dev/backends/raw_support_matrix.csv'
+    )
+    response.raise_for_status()
+    reader = csv.DictReader(response.text.splitlines())
+
+    data = [
+        {k: v if k == "Backends" else v == "True" for k, v in row.items()}
+        for row in reader
+    ]
+    for row in data:
+        row["FullOperation"] = row['Backends']
+        del row['Backends']
+        row['ShortOperation'] = row['FullOperation'].rsplit(".")[4]
+        row['OpCategory'] = row['FullOperation'].rsplit(".")[3]
+    return data
+
+
+ops_data = load_ops_data()
+
+
+class BackendCategory(Enum):
+    string = "STRING"
+    sqlalchemy = "SQLALCHEMY"
+    dataframe = "DATAFRAME"
+    sql = "SQL"
+
+
+BACKENDS = {
+    "bigquery": [BackendCategory.string, BackendCategory.sql],
+    "clickhouse": [BackendCategory.string, BackendCategory.sql],
+    'dask': [BackendCategory.dataframe],
+    "datafusion": [BackendCategory.dataframe],
+    "duckdb": [BackendCategory.sqlalchemy, BackendCategory.sql],
+    "impala": [BackendCategory.string, BackendCategory.sql],
+    "mssql": [BackendCategory.sqlalchemy, BackendCategory.sql],
+    "mysql": [BackendCategory.sqlalchemy, BackendCategory.sql],
+    "pandas": [BackendCategory.dataframe],
+    "polars": [BackendCategory.dataframe],
+    "postgres": [BackendCategory.sqlalchemy, BackendCategory.sql],
+    "pyspark": [BackendCategory.dataframe],
+    "snowflake": [BackendCategory.sqlalchemy, BackendCategory.sql],
+    "sqlite": [BackendCategory.sqlalchemy, BackendCategory.sql],
+    "trino": [BackendCategory.sqlalchemy, BackendCategory.sql],
+}
+
+
+def get_selected_backend_name():
+    all_categories_names = [cat.name for cat in iter(BackendCategory)]
+    selected_categories_names = st.sidebar.multiselect(
+        'Backend category',
+        options=sorted(all_categories_names),
+        default=None,
+    )
+    selected_categories = [
+        getattr(BackendCategory, name) for name in selected_categories_names
+    ]
+
+    selected_backend_names = {
+        backend_name
+        for backend_name, backend_types in BACKENDS.items()
+        if any(cat in backend_types for cat in selected_categories)
+    }
+    if not selected_backend_names:
+        selected_backend_names = list(BACKENDS.keys())
+    return selected_backend_names
+
+
+def get_selected_operation_categories():
+    all_ops_categories = {row['OpCategory'] for row in ops_data}
+    selected_ops_categories = st.sidebar.multiselect(
+        'Operation category',
+        options=sorted(all_ops_categories),
+        default=None,
+    )
+    if not selected_ops_categories:
+        return all_ops_categories
+    return selected_ops_categories
+
+
+current_backend_names = get_selected_backend_name()
+current_ops_categories = get_selected_operation_categories()
+hide_supported_by_all_backends = st.sidebar.selectbox(
+    'Operation compatibility',
+    ['Show all', 'Show supported by all backends', 'Hide supported by all backends'],
+    0,
+)
+show_full_ops_name = st.sidebar.checkbox('Show full operation name', False)
+
+df = pd.DataFrame(ops_data)
+if show_full_ops_name:
+    df = df.set_index("FullOperation")
+else:
+    df = df.set_index("ShortOperation")
+df = df.sort_index()
+
+# Show only selected operation
+df = df[df['OpCategory'].isin(current_ops_categories)]
+
+# Show only selected backend
+df = df[list(current_backend_names)]
+
+if hide_supported_by_all_backends == 'Show supported by all backends':
+    df = df[df.sum(1) == len(df.columns)]
+elif hide_supported_by_all_backends == 'Hide supported by all backends':
+    df = df[df.sum(1) != len(df.columns)]
+
+all_visible_ops_count = len(df.index)
+if all_visible_ops_count:
+    # Compute coverage
+    coverage = (
+        df.sum()
+        .sort_values(ascending=False)
+        .map(lambda n: f"{n} ({round(100 * n / all_visible_ops_count)}%)")
+        .to_frame(name="API Coverage")
+        .T
+    )
+
+    table = pd.concat(
+        [
+            coverage,
+            df.replace(
+                {
+                    True: "✔️",
+                    False: "✖",
+                }
+            ),
+        ]
+    )
+    st.table(table)
+else:
+    st.write("No data")
+
+with st.expander("Source code"):
+    st.code(Path(__file__).read_text())