From 4e16fea683ef7ac6ecee07c40e96e873bc42e9bb Mon Sep 17 00:00:00 2001
From: Seb Bacon <seb.bacon@gmail.com>
Date: Tue, 5 Mar 2019 14:47:47 +0000
Subject: [PATCH] Ensure we don't falsely claim to have cached something

Possibly related to #9
---
 ebmdatalab/__init__.py |  2 +-
 ebmdatalab/bq.py       |  6 ++++++
 ebmdatalab/test_bq.py  | 25 +++++++++++++++++++++++++
 3 files changed, 32 insertions(+), 1 deletion(-)

diff --git a/ebmdatalab/__init__.py b/ebmdatalab/__init__.py
index 718c4d1..a144f77 100644
--- a/ebmdatalab/__init__.py
+++ b/ebmdatalab/__init__.py
@@ -1,3 +1,3 @@
 """Package for ebmdatalab jupyter notebook stuff
 """
-__version__ = '0.0.7'
+__version__ = '0.0.8'
diff --git a/ebmdatalab/bq.py b/ebmdatalab/bq.py
index 0eff785..3e8a9ca 100644
--- a/ebmdatalab/bq.py
+++ b/ebmdatalab/bq.py
@@ -1,3 +1,4 @@
+import glob
 import os
 import re
 from hashlib import md5
@@ -43,6 +44,11 @@ def cached_read(sql,
     if use_cache and already_cached:
         df = pd.read_csv(csv_path)
     else:
+        old_fingerprint_files = glob.glob(
+            os.path.join(
+                csv_dir, "." + csv_filename + ".*.tmp"))
+        for f in old_fingerprint_files:
+            os.remove(f)
         with open(fingerprint_path, "w") as f:
             f.write("File created by {}".format(__file__))
         df = pd.read_gbq(sql, **defaults)
diff --git a/ebmdatalab/test_bq.py b/ebmdatalab/test_bq.py
index 23d5177..0101130 100644
--- a/ebmdatalab/test_bq.py
+++ b/ebmdatalab/test_bq.py
@@ -46,3 +46,28 @@ def test_cached_read(mock_read_gbq):
         # and now with `use_cache` param
         df = bq.cached_read(sql, csv_path=csv_file.name, use_cache=False)
         assert mock_read_gbq.call_count == 2
+
+
+def _check_cached_read(csv_file, mock_read, sql, expected):
+        mock_read.return_value = expected
+        df = bq.cached_read(sql, csv_path=csv_file.name)
+        assert str(df) == str(expected)
+
+
+@patch('ebmdatalab.bq.pd.read_gbq')
+def test_old_cache_markers_removed(mock_read_gbq):
+    with tempfile.NamedTemporaryFile() as csv_file:
+        # First, cause some sql to be cached
+        inputs_and_outputs = [
+            (
+                "select * from foobar",
+                DataFrame([{'a': 1}])
+            ),
+            (
+                "select * from foobar order by id",
+                DataFrame([{'a': 2}])
+            )
+        ]
+        _check_cached_read(csv_file, mock_read_gbq, *inputs_and_outputs[0])
+        _check_cached_read(csv_file, mock_read_gbq, *inputs_and_outputs[1])
+        _check_cached_read(csv_file, mock_read_gbq, *inputs_and_outputs[0])