Merge pull request #20 from mrchrisadams/ca-issue-19-double-quote-fal…

…lback

Add fallback exception when double quotes are used - see issue #19

.gitpod.yml

@@ -0,0 +1,12 @@
+# This configuration file was automatically generated by Gitpod.
+# Please adjust to your needs (see https://www.gitpod.io/docs/introduction/learn-gitpod/gitpod-yaml)
+# and commit this file to your remote git repository to share the goodness with others.
+
+# Learn more from ready-to-use templates: https://www.gitpod.io/docs/introduction/getting-started/quickstart
+
+tasks:
+  - name: Set up environment for development
+    init: |
+      python -m venv venv
+      source venv/bin/activate
+      pip install -e '.[test]'

datasette_parquet/exceptions.py

@@ -0,0 +1,23 @@
+class DoubleQuoteForLiteraValue(Exception):
+    """
+    DuckDB follows the SQL standard more closely than SQLite,
+    and as a result, is stricter about the use of double quotes
+    to wrap literal values.
+    Thrown when a literal value to compare against is wrapped in 
+    double quotes instead of single quotes.
+
+    For more info see the sqlite docs:
+    https://www.sqlite.org/quirks.html#double_quoted_string_literals_are_accepted
+    
+    """
+    def __init__(self, matches, message=None):
+        if message:
+            self.message = message
+        else:
+            self.message = (
+                "It looks like you are using a double quoted string "
+                f"for a value at: {matches[0]}. "
+                "To make this work with DuckDB, wrap it in single quoted "
+                "strings instead."
+        )
+        super().__init__(matches)

datasette_parquet/winging_it.py

@@ -1,9 +1,37 @@
 import time
+import re
+import typing
+
+from duckdb import BinderException 
+
 from .rewrite import rewrite, NO_OP_SQL
+from . import exceptions
+
+
+# a regular expression to find the literal wrapped by the double quoted string
+# GOOD for DuckDB, works fine:  
+# `WHERE column = 'some value'` 
+# 
+# BAD for DuckDB, triggers duckdb.BinderException:
+# `WHERE column = "some value"` 
+DOUBLE_QUOTES_AROUND_LITERALS_PATTERN = r'(?<![A-Za-z0-9_])"(?:[^"]*(?:"[^"]*)*[^"]*)"(?![A-Za-z0-9_])' # noqa
+
+def find_matching_double_quote_usage(ex: BinderException) -> typing.List[str]:
+    """
+    Accepts a BinderException from DuckDB, and returns a list of
+    matching strings in double quotes
+    """
+    # should return 'Binder Error: Referenced column "LITERAL_IN_DOUBLE_QUOTES" '
+    referenced_column_message = ex.args[0].split(' not found in FROM clause')[0]
+    pattern = DOUBLE_QUOTES_AROUND_LITERALS_PATTERN
+
+    # find the offending values wrapped in double quotes
+    matches = re.findall(pattern, referenced_column_message)
+    return matches
+
 
 # A collection of classes to provide a facade that mimics the sqlite3 DB-API
 # interface.
-
 class Row:
     def __init__(self, columns, tpl):
         self.columns = columns
@@ -62,7 +90,16 @@ def execute(self, sql, parameters=None):
 
         #print('## params={} sql={}'.format(parameters, sql))
         t = time.time()
-        rv = self.cursor.execute(sql, parameters)
+        try:
+            rv = self.cursor.execute(sql, parameters)
+        except BinderException as ex:
+            matches = find_matching_double_quote_usage(ex)
+            if matches:
+                raise exceptions.DoubleQuoteForLiteraValue(matches)
+            else:
+                # continue raising the original BinderException
+                raise
+
         #print('took {}'.format(time.time() - t))
         return rv
 
@@ -128,8 +165,16 @@ def execute(self, sql, parameters=None):
         #print('! rewritten sql={}'.format(sql))
         sql, parameters = fixup_params(sql, parameters)
         #print('!! params={} sql={}'.format(parameters, sql))
-        rv = self.conn.execute(sql, parameters)
-
+        try:
+            rv = self.conn.execute(sql, parameters)
+        except BinderException as ex:
+            matches = find_matching_double_quote_usage(ex)
+            if matches:
+                raise exceptions.DoubleQuoteForLiteraValue(matches)
+            else:
+                # continue raising the original BinderException
+                raise
+
         return ProxyCursor(self.conn, rv)
 
     def fetchall(self):

tests/test_parquet.py

@@ -3,6 +3,7 @@
 import pytest
 import duckdb
 from datasette_parquet.winging_it import ProxyConnection
+from datasette_parquet import exceptions
 
 @pytest.fixture(scope="session")
 def datasette():
@@ -71,3 +72,16 @@ def test_fetchone():
     conn = ProxyConnection(raw_conn)
     fetched = conn.execute('SELECT 1 AS col').fetchone()
     assert fetched['col'] == 1
+
+
+@pytest.mark.asyncio
+def test_catch_double_quote_usage_for_literal(datasette):
+
+    raw_conn = duckdb.connect()
+    conn = ProxyConnection(raw_conn)
+
+    # try reading the parquet file in trove/userdata1.parquet
+    explodey_string_with_double_quotes = 'SELECT * from "./trove/userdata1.parquet" WHERE first_name = "Amanda"'
+
+    with pytest.raises(exceptions.DoubleQuoteForLiteraValue):
+        result = conn.execute(explodey_string_with_double_quotes).fetchall()