pig: first rev of join-cdx-sha1 script

2 files changed, 48 insertions, 0 deletions

diff --git a/pig/tests/files/example.sha1b32 b/pig/tests/files/example.sha1b32
new file mode 100644
index 0000000..20a1357
--- /dev/null
+++ b/pig/tests/files/example.sha1b32
@@ -0,0 +1,4 @@
+EJWYVOPONJRARK7SGG6COFRN7CSTHROY
+V32E3CCO7NMI2M4OHLKG73DXD72LR4B2
+3I42H3S6NNFQ2MSVX7XZKYAYSCX5QBYJ
+E3WSNQ7JAFOW7N3ZJ6GLV27T52T25JDK
diff --git a/pig/tests/test_join_cdx.py b/pig/tests/test_join_cdx.py
new file mode 100644
index 0000000..e6eca6a
--- /dev/null
+++ b/pig/tests/test_join_cdx.py
@@ -0,0 +1,44 @@
+
+import os
+import unittest
+import tempfile
+import subprocess
+from pighelper import PigTestHelper, count_lines
+
+class TestJoinCDXSha1(PigTestHelper):
+
+    def run_pig_join(self, script_path, cdx_file, digest_file, **kwargs):
+        """Convenience helper around run_pig().
+        
+        INPUT parameter is set to in_file.
+        OUTPUT parameter is set to a random file.
+        Any keyword args are passed as parameters.
+        """
+
+        pargs = []
+        for key, value in kwargs.items():
+            pargs.append('-p')
+            pargs.append('{}={}'.format(key, value))
+
+        out_file = tempfile.mktemp(dir=self._tmpdir)
+        params = [
+            '-f', script_path,
+            '-p', 'INPUT_CDX={}'.format(cdx_file),
+            '-p', 'INPUT_DIGEST={}'.format(digest_file),
+            '-p', 'OUTPUT={}'.format(out_file),
+            ] + pargs
+        status = self.run_pig_raw(params)
+        assert status.returncode == 0
+        # Capture all the part-r-* files
+        print("out_file: {}".format(out_file))
+        subprocess.run("/bin/ls -la {}/part-*".format(out_file), shell=True)
+        sub = subprocess.run("/bin/cat {}/part-*".format(out_file), stdout=subprocess.PIPE, shell=True)
+        out = sub.stdout.decode('utf-8')
+        print(out)
+        return out
+
+    # TODO: helper to verify that output matches an expected file
+
+    def test_thing(self):
+        r = self.run_pig_join("join-cdx-sha1.pig", "tests/files/example.cdx", "tests/files/example.sha1b32")
+        assert count_lines(r) == 4