def test_valid_database(self):
"""Tests a good database to make sure there are no false positives."""
good_db = Database(database_name="good_db")
table1 = Table(
table_name="table1",
primary_key=["col1", "col2"],
shard_key=ShardKey(shard_keys="col1", number_shards=128),
)
table1.add_column(Column(column_name="col1", column_type="INT"))
table1.add_column(Column(column_name="col2", column_type="INT"))
good_db.add_table(table1)
table2 = Table(
table_name="table2",
primary_key=["col3", "col4"],
shard_key=ShardKey(shard_keys="col3", number_shards=128),
)
table2.add_foreign_key(
from_keys=["col3", "col4"],
to_table="table1",
to_keys=["col1", "col2"],
)
table2.add_column(Column(column_name="col3", column_type="INT"))
table2.add_column(Column(column_name="col4", column_type="INT"))
good_db.add_table(table2)
dv = DatabaseValidator(good_db)
results = dv.validate()
self.assertTrue(results.is_valid)
self.assertEqual([], results.issues)
def get_complex_db():
"""
Returns a more complex database with two tables and keys for testing.
:return: Database with two tables and keys.
:rtype: Database
"""
database = Database(database_name="database2")
table1 = Table(table_name="table1",
primary_key="col1",
shard_key=ShardKey("col1", 128))
table1.add_column(Column(column_name="col1", column_type="INT"))
table1.add_column(Column(column_name="Col2", column_type="DOUBLE"))
table1.add_column(Column(column_name="COL3", column_type="FLOAT"))
database.add_table(table1)
table2 = Table(table_name="table2",
primary_key=["col4", "Col5"],
shard_key=ShardKey(["col4", "Col5"], 96))
table2.add_column(Column(column_name="col4", column_type="VARCHAR(0)"))
table2.add_column(Column(column_name="Col5", column_type="DATE"))
table2.add_column(Column(column_name="COL6", column_type="BOOL"))
database.add_table(table2)
table2.add_foreign_key(from_keys="Col5",
to_table="table1",
to_keys="COL3")
table1.add_relationship(
to_table="table2",
conditions='("table1"."col1" == "table2."COL6")')
return database
def test_create_excel(self):
"""Test writing to Excel. Only test is existance. Checks shoudl be made for validity."""
database = Database(database_name="xdb")
table = Table(table_name="table1",
schema_name="s1",
primary_key="column_1",
shard_key=ShardKey("column_1", 128))
table.add_column(Column(column_name="column_1", column_type="INT"))
table.add_column(Column(column_name="column_2", column_type="DOUBLE"))
table.add_column(Column(column_name="column_3", column_type="FLOAT"))
database.add_table(table)
table = Table(table_name="table2",
schema_name="s1",
primary_key="column_1")
table.add_column(Column(column_name="column_1", column_type="INT"))
table.add_column(Column(column_name="column_2",
column_type="DATETIME"))
table.add_column(Column(column_name="column_3", column_type="BOOL"))
table.add_column(Column(column_name="column_4", column_type="DOUBLE"))
table.add_foreign_key(from_keys="column_1",
to_table="table_1",
to_keys="column_1")
table.add_relationship(to_table="table1",
conditions="table2.column_4 = table1.column_2")
database.add_table(table)
writer = XLSWriter()
writer.write_database(database, "test_excel")
def test_change_column(self):
"""Tests adding / dropping a column from a table."""
dc = DDLCompare()
db1 = Database(database_name="database1")
db2 = Database(database_name="database2")
t1 = Table(table_name="table1")
t1.add_column(column=Column(column_name="column1", column_type="INT"))
db1.add_table(t1)
t2 = Table(table_name="table1")
t2.add_column(column=Column(column_name="column1", column_type="FLOAT"))
db2.add_table(t2)
diff1, diff2 = dc.compare_databases(db1=db1, db2=db2)
self.assertTrue(type(diff1[0] is ColumnModifiedDifference))
self.assertEqual(diff1[0].table_name, "table1")
self.assertEqual(diff1[0].column.column_name, "column1")
self.assertEqual(diff1[0].column.column_type, "FLOAT")
self.assertTrue(type(diff2[0] is ColumnModifiedDifference))
self.assertEqual(diff2[0].table_name, "table1")
self.assertEqual(diff2[0].column.column_name, "column1")
self.assertEqual(diff2[0].column.column_type, "INT")
def _read_tables_from_workbook(self):
"""
Reads the databases and tables from Excel. These are used to populate from the remaining sheets.
"""
# "Tables": ["Database", "Schema", "Table", "Updated", "Update Type", "# Rows", "# Columns",
# "Primary Key", "Shard Key", "# Shards", "RLS Column"],
table_sheet = self.workbook.sheet_by_name("Tables")
indices = self.indices["Tables"]
for row_count in range(1, table_sheet.nrows):
row = table_sheet.row_values(rowx=row_count, start_colx=0)
database_name = row[indices["Database"]]
database = self.databases.get(database_name, None)
if database is None:
database = Database(database_name=database_name)
self.databases[database_name] = database
pk = row[indices["Primary Key"]].strip()
if pk == "":
pk = None
else:
pk = [x.strip() for x in pk.split(",")]
sk_name = row[indices["Shard Key"]].strip()
sk_nbr_shards = row[indices["# Shards"]]
if (sk_name == "" and sk_nbr_shards != "") or (
sk_name != "" and sk_nbr_shards == ""
):
eprint(
"ERROR: %s need to provide both a shard key name and number of shards."
% row[indices["Table"]]
)
if sk_name == "":
sk = None
else:
sk = [x.strip() for x in sk_name.split(",")]
shard_key = None
if sk_name != "" and sk_nbr_shards != "":
shard_key = ShardKey(
shard_keys=sk, number_shards=sk_nbr_shards
)
table = Table(
table_name=row[indices["Table"]],
schema_name=row[indices["Schema"]],
primary_key=pk,
shard_key=None,
)
database.add_table(table)
def get_simple_db():
"""
Returns a simple database with one table for testing.
:return: Database with one table.
:rtype: Database
"""
database = Database(database_name="database1")
table = Table(table_name="table1")
table.add_column(Column(column_name="col1", column_type="INT"))
table.add_column(Column(column_name="Col2", column_type="DOUBLE"))
table.add_column(Column(column_name="COL3", column_type="FLOAT"))
database.add_table(table)
return database
def test_new_and_drop_table(self):
"""Tests adding / dropping a table. It's added in one and dropped in the other."""
dc = DDLCompare()
db1 = Database(database_name="database1")
db2 = Database(database_name="database2")
db1.add_table(Table(table_name="table_from_1"))
diff1, diff2 = dc.compare_databases(db1=db1, db2=db2)
self.assertTrue(type(diff1[0]) is TableDroppedDifference)
self.assertEqual(diff1[0].table_name, "table_from_1")
self.assertTrue(type(diff2[0]) is TableCreatedDifference)
self.assertEqual(diff2[0].table_name, "table_from_1")
def test_add_and_drop_fk(self):
"""Tests adding / dropping a column from a table."""
dc = DDLCompare()
db1 = Database(database_name="database1")
db2 = Database(database_name="database2")
t1 = Table(table_name="table1", primary_key="column1")
db1.add_table(t1)
t2 = Table(table_name="table1")
db2.add_table(t2)
diff1, diff2 = dc.compare_databases(db1=db1, db2=db2)
self.assertTrue(type(diff1[0] is PrimaryKeyDroppedDifference))
self.assertEqual(diff1[0].table_name, "table1")
self.assertTrue(type(diff2[0] is PrimaryKeyAddedDifference))
self.assertEqual(diff2[0].table_name, "table1")
def test_add_and_drop_rel(self):
"""Tests adding / dropping a column from a table."""
dc = DDLCompare()
db1 = Database(database_name="database1")
db2 = Database(database_name="database2")
t1 = Table(table_name="table1")
t1.add_relationship(relationship=GenericRelationship(from_table="table_1", to_table="table_2",
conditions="table1.col1 = table2.col2"))
db1.add_table(t1)
t2 = Table(table_name="table1")
db2.add_table(t2)
diff1, diff2 = dc.compare_databases(db1=db1, db2=db2)
self.assertTrue(type(diff1[0] is GenericRelationshipDroppedDifference))
self.assertEqual(diff1[0].table_name, "table1")
self.assertTrue(type(diff2[0] is GenericRelationshipAddedDifference))
self.assertEqual(diff2[0].table_name, "table1")
def test_with_csvfile(self):
"""test the tsload writer when the csv exists"""
# todo Create the csv file.
database = Database(database_name="xdb")
table = Table(
table_name="table1",
schema_name="s1",
primary_key="column_1",
shard_key=ShardKey("column_1", 128),
)
table.add_column(Column(column_name="column_1", column_type="INT"))
table.add_column(Column(column_name="column_2", column_type="DOUBLE"))
table.add_column(Column(column_name="column_3", column_type="FLOAT"))
table.add_column(Column(column_name="column_3", column_type="DATE"))
database.add_table(table)
table = Table(
table_name="table2",
schema_name="s1",
primary_key="column_1",
shard_key=ShardKey("column_1", 128),
)
table.add_column(Column(column_name="column_1", column_type="INT"))
table.add_column(Column(column_name="column_2", column_type="FLOAT"))
table.add_column(Column(column_name="column_3", column_type="DOUBLE"))
database.add_table(table)
table = Table(
table_name="table3",
schema_name="s1",
primary_key="column_1",
shard_key=ShardKey("column_1", 128),
)
table.add_column(Column(column_name="column_1", column_type="INT"))
table.add_column(Column(column_name="column_2", column_type="FLOAT"))
table.add_column(Column(column_name="column_3", column_type="VARCHAR"))
database.add_table(table)
tsload_writer = TsloadWriter()
tsload_writer.write_tsloadcommand(database, "tsloadwriter_test")
with open("tsloadwriter_test", "r") as infile:
line = infile.readline()
self.assertTrue(line.startswith("tsload "))
self.assertTrue(line.index('--target_database "xdb"') > 0)
self.assertTrue(line.index('--target_schema "s1"'))
def create_test_database():
"""Creates a database for testing."""
database = Database("database1")
database.add_table(Table(table_name="table1", schema_name="schema1"))
database.add_table(Table(table_name="table2"))
return database
class DDLParser(object):
"""
Parses DDL from various formats and creates a DataModel object that can be used for writing data.
The following assumptions are made about the DDL being read:
* CREATE TABLE occur together on a single line, not split across lines.
* CREATE TABLE statements will not occur inside of a comment block.
* Delimiters, such as commas, will not be part of the table or column name.
* Comment characters, such as #, --, or /* */ will not be part of a column name.
* CREATE TABLE will have (....) with no embedded, unbalanced parentheses.
"""
# TODO: capture primary keys, foreign keys, and relationships.
def __init__(
self,
database_name,
schema_name=DatamodelConstants.DEFAULT_SCHEMA,
parse_keys=False,
):
"""
Creates a new DDL parser.
:param database_name: Name of the database to create.
:type database_name: str
:param schema_name: Name of the schema if not using the default.
:param parse_keys: If true, the parser will attempt to parse keys as well.
:type parse_keys: bool
:type schema_name: str
"""
self.schema_name = schema_name
self.database = Database(database_name=database_name)
self.parse_keys = parse_keys
def parse_ddl(self, filename):
"""
Parsed DDL from a stream and returns a populated Database.
:param filename: Name of the file to read from.
:return: A Database object.
:rtype: Database
"""
if filename is None:
stream = open(sys.stdin, "r")
else:
stream = open(filename, "r")
# First read the entire input into memory. This will allow multiple passes through the data.
input_ddl = []
for line in stream:
input_ddl.append(line)
self.parse_tables(input_ddl)
if self.parse_keys:
self.parse_primary_keys(input_ddl)
return self.database
def parse_tables(self, input_ddl):
"""
Parses the input DDL to convert to a database model.
:param input_ddl: The DDL to convert.
:type input_ddl: list of str
"""
creating = False
buff = ""
for line in input_ddl:
l = self.clean_line(line)
if not creating: # looking for CREATE TABLE statement.
if l.lower().find("create table") >= 0:
creating = True
buff = l
if self.is_complete_create(buff):
self.parse_create_table(buff)
buff = ""
creating = False
else: # looking for the end of a create table.
buff += l
if self.is_complete_create(buff):
self.parse_create_table(buff)
buff = ""
creating = False
@staticmethod
def is_complete_create(buff):
"""
Returns true if the number of open and close parentheses match.
:param buff: The buffer being read.
:return: str
"""
nbr_open = buff.count("(")
nbr_close = buff.count(")")
return nbr_open > 0 and nbr_open == nbr_close
def parse_create_table(self, buff):
"""
Parses a create table statement.
:param buff: The buffer read in.
:type buff: str
:return:
"""
buff = buff.replace("[", '"').replace(
"]", '"'
) # for SQL Server quotes
table_name = self.get_table_name(buff)
columns = self.get_columns(buff)
table = Table(table_name=table_name, schema_name=self.schema_name)
table.add_columns(columns)
self.database.add_table(table)
def get_table_name(self, buff):
"""
Gets the table name from the buffer.
:param buff: The line with the create details.
:type buff: str
:return: The name of the table.
:rtype: str
"""
# The table name (and maybe a schema) are before the opening (
tn = buff[0:buff.find("(")].rstrip()
# split on spaces and assume last one is table name (and maybe schema)
tn = tn.split(" ")[-1]
tn = tn.split(".")[-1]
tn = self.strip_quotes(tn)
return tn
@staticmethod
def strip_quotes(line):
"""
Strips off any quotes in the given line.
:param line: The line to strip quotes from.
:type line: str
:return: The line without quotes.
:rtype: str
"""
return line.replace("'", "").replace("`", "").replace('"', "")
def get_columns(self, buff):
"""
Get the columns from the table statement.
:param buff: The buffer with the create details.
:type buff: str
:return: A list of Columns
:rtype: list
"""
# The fields will be between the ( ).
columns = []
buff = buff[buff.find("(") + 1:buff.rfind(")")].strip()
# think all DBs use commas for field separators
# need to find the commas that are not inside of parents.
field_buff = ""
open_paren = False
raw_fields = []
for c in buff:
if open_paren:
field_buff += c
if c == ")":
open_paren = False
elif c == "(":
field_buff += c
open_paren = True
else:
if c == ",":
raw_fields.append(field_buff)
field_buff = ""
else:
field_buff += c
if field_buff != "":
raw_fields.append(field_buff)
for rf in raw_fields:
rfl = rf.lower()
# ignore key declarations.
if "key " in rfl:
continue
had_quote = False
if rfl[0] in "\"'`": # should be a quote or letter
had_quote = True
name = rf[1:rf.find(rf[0], 1)]
else:
name = rf[0:rf.find(" ")]
# The type comes after the name and goes up to the first of a
# space, close paren, or comma. Assuming no space in type.
start_idx = len(name) + (
3 if had_quote else 1
) # extra 1 for space
if rfl.find(")") > 0: # type with ()
data_type = rf[start_idx:rf.find(")") + 1]
else:
# either next space or comma.
space_end_idx = rf.find(" ", start_idx)
comma_end_idx = rf.find(",", start_idx)
if space_end_idx == -1: # not found
if comma_end_idx == -1: # neither found
end_idx = len(rf) # end of line
else:
end_idx = comma_end_idx
elif comma_end_idx == -1:
end_idx = space_end_idx
else:
end_idx = min(space_end_idx, comma_end_idx)
data_type = rf[start_idx:end_idx]
# print (" adding %s as %s" % (name, data_type))
columns.append(
Column(
column_name=name, column_type=self.convert_type(data_type)
)
)
return columns
@staticmethod
def convert_type(data_type):
"""
Converts data types from other databases to ThoughtSpot types.
:param data_type: The datatype to convert.
:type data_type: str
:return: A ThoughtSpot data type.
:rtype: str
"""
if ")" in data_type:
t = data_type[0:data_type.find(")") + 1]
elif " " in data_type:
t = data_type[0:data_type.find(" ") + 1]
else:
t = data_type
t = t.lower()
if "int" in t:
new_t = "BIGINT"
elif "rowversion" in t: # MS type
new_t = "INT"
elif "uniqueidentifier" in t: # Oracle type
new_t = "VARCHAR(0)"
elif "serial" in t: # serial index, Oracle and others
new_t = "INT"
elif "bit" in t:
new_t = "BOOL"
elif "blob" in t or "binary" in t:
new_t = "UNKNOWN"
elif "number" in t: # support for NUMBER(1), NUMBER(1,1)
if ")" in t:
numsize = t[t.find("(") + 1:t.find(")")]
if "," in numsize:
first_num, second_num = numsize.split(",")
if second_num.strip() == "0":
if first_num == "*" or int(
first_num
) > 9: # Support Oracle Number(*,n)
new_t = "BIGINT"
else:
new_t = "INT"
else:
new_t = "DOUBLE"
else:
new_t = "INT"
else:
new_t = "BIGINT"
elif "decimal" in t or "numeric" in t or "float" in t or "double" in t or "money" in t or "real" in t:
new_t = "DOUBLE"
elif "datetime" in t:
new_t = "DATETIME"
elif "timestamp" in t:
new_t = "DATETIME"
elif "time" in t:
new_t = "TIME"
elif "date" in t:
new_t = "DATE"
elif "bool" in t:
new_t = "BOOL"
elif "text" in t:
new_t = "VARCHAR(0)"
elif "long" in t: # Oracle variable type
new_t = "VARCHAR(0)"
elif "enum" in t:
new_t = "VARCHAR(0)"
elif "xml" in t:
new_t = "VARCHAR(0)"
elif "char" in t:
new_t = "VARCHAR(0)"
# nbytes = 0
# if ')' in t:
# nbytes = t[t.find('(') + 1:t.find(')')]
# nbytes = re.sub("[^0-9]", "", nbytes)
# if nbytes == "":
# nbytes = 0
# new_t = "VARCHAR(%s)" % nbytes
else:
new_t = "UNKNOWN"
return new_t
def parse_primary_keys(self, input_ddl):
"""
Parses primary keys (and shard keys for TQL.
:param input_ddl: The input DDL to parse from.
:type input_ddl: list of str
"""
# read through lines until a CREATE TABLE or ALTER TABLE is found.
# next look for either a new CREATE TABLE, a PRIMARY KEY, or PARTITION BY HASH
# add the primary key or partition
# TODO - get this to work.
create_or_update = False
buff = ""
for line in input_ddl:
l = self.clean_line(line)
if not create_or_update: # looking for CREATE TABLE or UPDATE TABLE statement.
if l.lower().find("create table") >= 0 or l.lower().find(
"update table"
) >= 0:
create_or_update = True
buff = l
if self.is_complete_create(buff):
self.parse_create_table(buff)
buff = ""
create_or_update = False
else: # looking for the end of a create table.
buff += l
if self.is_complete_create(buff):
self.parse_create_table(buff)
buff = ""
create_or_update = False
pass
@staticmethod
def clean_line(line):
"""
Removes unwanted characters from the input line.
:param line: The line to clean up.
:type line: str
:return: The cleaned up line.
:rtype: str
"""
l = line.strip()
l = re.sub(" +", " ", l)
l = re.sub("\t+", " ", l)
return l