Add functions for creating directories, dimensions, etc.

Refactor unit tests to use above functions.

snakepit/create.py

@@ -1,6 +1,6 @@
 import sqlalchemy as sq
 
-from snakepit import hive
+from snakepit import hive, directory
 
 def create_hive(hive_uri):
     """
@@ -21,3 +21,111 @@ def create_hive(hive_uri):
         table.tometadata(hive_metadata)
     hive_metadata.create_all()
     return hive_metadata
+
+
+def create_primary_index(
+    directory_uri,
+    dimension_name,
+    ):
+    """
+    Create a primary index for C{dimension_name} at C{directory_uri}.
+    """
+    directory_metadata = sq.MetaData()
+    directory_metadata.bind = sq.create_engine(
+        directory_uri,
+        strategy='threadlocal',
+        )
+    table = directory.dynamic_table(
+        table=directory.metadata.tables['hive_primary_DIMENSION'],
+        metadata=directory_metadata,
+        name='hive_primary_%s' % dimension_name,
+        )
+    directory_metadata.create_all()
+    return directory_metadata
+
+
+class DimensionExistsError(Exception):
+    """Dimension exists already"""
+
+    def __str__(self):
+        return ': '.join([self.__doc__]+list(self.args))
+
+def create_dimension(hive_metadata, dimension_name, directory_uri):
+    """
+    Create a dimension with C{dimension_name} at C{hive_metadata},
+    where the directory index is stored at C{directory_uri}.
+
+    Directory index must be set up before calling this function.
+
+    @raise DimensionExistsError: a dimension with that name exists
+    already in this hive
+    """
+    t = hive_metadata.tables['partition_dimension_metadata']
+    try:
+        r = t.insert().execute(
+            name=dimension_name,
+            index_uri=directory_uri,
+            db_type=0, # TODO
+            )
+    except sq.exceptions.SQLError, e:
+        # sqlalchemy 0.3.x is hiding details and not providing a
+        # db-independent abstraction for what the error actually
+        # was, so we need to resort to kludges
+
+        # only catch sq.exceptions.IntegrityError when it's safe
+        # to depend on sqlalchemy 0.4
+
+        # http://www.sqlalchemy.org/trac/ticket/706
+        if 'IntegrityError' in unicode(e):
+            raise DimensionExistsError(repr(dimension_name))
+        else:
+            raise
+
+    (dimension_id,) = r.last_inserted_ids()
+    r.close()
+    return dimension_id
+
+class NodeExistsError(Exception):
+    """Node exists already"""
+
+    def __str__(self):
+        return ': '.join([self.__doc__]+list(self.args))
+
+def create_node(hive_metadata, dimension_id, node_name, node_uri):
+    """
+    Create a node with in dimension having C{dimension_id} with
+    C{node_name} at C{hive_metadata}, where the records are stored at
+    C{node_uri}.
+
+    Node must be set up before calling this function.
+
+    @raise NoSuchDimensionError: no such dimension found
+
+    @raise NodeExistsError: a node with that name exists
+    already in this hive
+    """
+    t = hive_metadata.tables['node_metadata']
+    try:
+        r = t.insert().execute(
+            partition_dimension_id=dimension_id,
+            name=node_name,
+            uri=node_uri,
+            read_only=False, # TODO
+            )
+    except sq.exceptions.SQLError, e:
+        # sqlalchemy 0.3.x is hiding details and not providing a
+        # db-independent abstraction for what the error actually
+        # was, so we need to resort to kludges
+
+        # only catch sq.exceptions.IntegrityError when it's safe
+        # to depend on sqlalchemy 0.4
+
+        # http://www.sqlalchemy.org/trac/ticket/706
+        if 'IntegrityError' in unicode(e):
+            raise NodeExistsError(repr(node_name))
+        else:
+            raise
+
+    (node_id,) = r.last_inserted_ids()
+    r.close()
+    return node_id

snakepit/hive.py

@@ -22,13 +22,19 @@
     sq.Column('uri', sq.String(255), nullable=False),
     # this is int in HiveConfigurationSchema.java
     sq.Column('read_only', sq.Boolean),
+
+    # TODO what's the scope of the name? I guess you could
+    # say name is hostname, and same machine can serve two
+    # dimensions -- then name isn't unique alone, only in
+    # combination with partition_dimension_id
+    sq.UniqueConstraint('partition_dimension_id', 'name'),
     )
 
 partition_dimension_metadata = sq.Table(
     'partition_dimension_metadata',
     metadata,
     sq.Column('id', sq.Integer, primary_key=True),
-    sq.Column('name', sq.String(64), nullable=False),
+    sq.Column('name', sq.String(64), nullable=False, unique=True),
     sq.Column('index_uri', sq.String(255), nullable=False),
     # TODO wth is this used?
     sq.Column('db_type', sq.String(64), nullable=False),