Merge pull request #264 from efiop/master

Misc fixes and improvements

dvc/command/merge.py

@@ -10,37 +10,37 @@ class CmdMerge(CmdBase):
         super(CmdMerge, self).__init__(settings)
 
     def print_info(self, targets):
-        for fname in targets:
+        for item in targets:
             Logger.info('Restored original data after merge:')
-            Logger.info(' {}'.format(fname))
+            Logger.info(' {}'.format(item.data.relative))
 
     def collect_targets(self):
         targets = []
         flist = self.git.get_last_merge_changed_files()
-        for fname in flist:
-            try:
-                item = self.settings.path_factory.data_item(fname)
-            except NotInDataDirError:
-                continue
+        items = self.settings.path_factory.to_data_items(flist)[0]
 
+        for item in items:
             try:
                 state = StateFile.load(item, self.git)
             except Exception as ex:
-                Logger.error('Failed to load state file for {}'.format(fname), exc_info=True)
+                Logger.error('Failed to load state file for {}'.format(item.data.relative), exc_info=True)
                 return None
 
             if not state.command == StateFile.COMMAND_IMPORT_FILE:
                 continue
 
-            targets.append(fname)
+            targets.append(item)
 
         return targets
 
     def checkout_targets(self, targets):
-        for fname in targets:
-            self.git.checkout_file_before_last_merge(fname)
+        data = []
+        for item in targets:
+            self.git.checkout_file_before_last_merge(item.data.relative)
+            self.git.checkout_file_before_last_merge(item.state.relative)
+            data.append(item.data.relative)
 
-        msg = 'DVC merge files: {}'.format(' '.join(targets))
+        msg = 'DVC merge files: {}'.format(' '.join(data))
         self.commit_if_needed(msg)
 
     def run(self):

dvc/command/repro.py

@@ -78,7 +78,7 @@ class CmdRepro(CmdRun):
 
         for data_item in data_item_list:
             try:
-                target_commit = self.git.get_target_commit(data_item.data.relative)
+                target_commit = self.git.get_target_commit(data_item.state.relative)
                 if target_commit is None:
                     msg = 'Data item "{}" cannot be reproduced: cannot obtain commit hashsum'
                     Logger.warn(msg.format(data_item.data.relative))

dvc/state_file.py

@@ -113,7 +113,13 @@ class StateFile(object):
 
     @staticmethod
     def parse_target_metric(file_name):
-        lines = open(file_name).readlines(2)
+        fd = open(file_name, 'r')
+
+        try:
+            lines = fd.readlines(2)
+        except Exception as e:
+            raise StateFileError('Unable to parse target file')
+
         if len(lines) != 1:
             raise StateFileError('file contains more then one line')
 

functest/run.sh

@@ -8,6 +8,7 @@ TESTS+=" test-init.sh"
 TESTS+=" test-import.sh"
 TESTS+=" test-sync.sh"
 TESTS+=" test-repro-code.sh"
+TESTS+=" test-merge.sh"
 
 for t in $TESTS; do
 	rm -rf $TEST_REPO

functest/test-merge.sh

@@ -0,0 +1,55 @@
+#!/bin/bash
+
+set -e
+
+source common.sh
+
+dvc_create_repo
+
+# Adding changes to the first branch
+git checkout -b input_50K
+echo -e "#!/bin/bash\nset -x\nset -e\ncp \$1 \$2" > code/code1.sh
+chmod +x code/code1.sh
+cp code/code1.sh code/code2.sh
+cp code/code1.sh code/code3.sh
+cp code/code1.sh code/code4.sh
+git add code
+git commit -s -m"add code"
+
+dvc import https://s3-us-west-2.amazonaws.com/dvc-share/so/50K/Posts.xml.tgz data/data
+
+ORIG_DATA=$(readlink -- data/data)
+
+dvc run ./code/code1.sh data/data data/data1
+dvc run ./code/code2.sh data/data1 data/data2
+dvc run ./code/code3.sh data/data2 data/data3
+dvc run ./code/code4.sh data/data3 data/data4
+
+dvc config Global.Target data/data4
+git commit -am 'Set Target'
+
+# Create new branch to work on small data set
+git checkout -b input_10K
+dvc remove data/data
+dvc import https://s3-us-west-2.amazonaws.com/dvc-share/so/10K/Posts.xml.tgz data/data
+dvc repro
+
+echo -e "#!/bin/bash\n\n\n\ncp \$1 \$2" > code/code1.sh
+git commit -am"change code1.sh"
+dvc repro
+
+# Merge input_10K into input_50K
+git checkout input_50K
+git merge input_10K
+# Notice we're running 'dvc merge' on input_50K branch. This
+# way dvc knows to favor data files from current branch.
+dvc merge
+
+# Verify that data and respective state file are restored properly
+if [ "$(readlink -- "data/data")" != "$ORIG_DATA" ]; then
+	dvc_fail
+fi
+
+grep $(basename "$ORIG_DATA") .dvc/state/data.state || dvc_fail
+
+dvc_pass