-from collections import OrderedDict
-import sqlparse
import tables
+import datasette_connectors as dc
+from .utils import parse_sql
-_connector_type = 'pytables'
-def inspect(path):
- "Open file and return tables info"
- h5tables = {}
- views = []
- h5file = tables.open_file(path)
+class PyTablesConnection(dc.Connection):
+ def __init__(self, path, connector):
+ super().__init__(path, connector)
+ self.h5file = tables.open_file(path)
+
- for table in filter(lambda node: not(isinstance(node, tables.group.Group)), h5file):
- colnames = []
+class PyTablesConnector(dc.Connector):
+ connector_type = 'pytables'
+ connection_class = PyTablesConnection
+
+ operators = {
+ 'eq': '==',
+ 'neq': '!=',
+ 'gt': '>',
+ 'gte': '>=',
+ 'lt': '<',
+ 'lte': '<=',
+ 'and': '&',
+ 'or': '|',
+ 'binary_and': '&',
+ 'binary_or': '|',
+ }
+
+ def table_names(self):
+ return [
+ node._v_pathname
+ for node in self.conn.h5file
+ if not(isinstance(node, tables.group.Group))
+ ]
+
+ def table_count(self, table_name):
+ table = self.conn.h5file.get_node(table_name)
+ return int(table.nrows)
+
+ def table_info(self, table_name):
+ table = self.conn.h5file.get_node(table_name)
+ colnames = ['value']
if isinstance(table, tables.table.Table):
colnames = table.colnames
- h5tables[table._v_pathname] = {
- 'name': table._v_pathname,
- 'columns': colnames,
- 'primary_keys': [],
- 'count': table.nrows,
- 'label_column': None,
- 'hidden': False,
- 'fts_table': None,
- 'foreign_keys': {'incoming': [], 'outgoing': []},
- }
-
- h5file.close()
- return h5tables, views, _connector_type
-
-def _parse_sql(sql):
- parsed = sqlparse.parse(sql)
- stmt = parsed[0]
- parsed_sql = {}
- current_keyword = ""
- for token in stmt.tokens:
- if token.is_keyword:
- if current_keyword in parsed_sql and parsed_sql[current_keyword] == '':
- # Check composed keywords like 'order by'
- del parsed_sql[current_keyword]
- current_keyword += " " + str(token)
- else:
- current_keyword = str(token)
- parsed_sql[current_keyword] = ""
- elif type(token) is sqlparse.sql.Where:
- parsed_sql['where'] = token
- else:
- if not token.is_whitespace:
- parsed_sql[current_keyword] += str(token)
- return parsed_sql
-
-_operators = {
- '=': '==',
-}
-
-def _translate_condition(table, condition, params):
- field = condition.left.get_real_name()
-
- operator = list(filter(lambda t: t.ttype == sqlparse.tokens.Comparison, condition.tokens))[0]
- if operator.value in _operators:
- operator = _operators[operator.value]
- else:
- operator = operator.value
-
- value = condition.right.value
- if value.startswith(':'):
- # Value is a parameters
- value = value[1:]
- if value in params:
+ return [
+ {
+ 'idx': idx,
+ 'name': colname,
+ 'primary_key': False,
+ }
+ for idx, colname in enumerate(colnames)
+ ]
+
+ def hidden_table_names(self):
+ return []
+
+ def detect_spatialite(self):
+ return False
+
+ def view_names(self):
+ return []
+
+ def detect_fts(self, table_name):
+ return False
+
+ def foreign_keys(self, table_name):
+ return []
+
+ def execute(
+ self,
+ sql,
+ params=None,
+ truncate=False,
+ custom_time_limit=None,
+ page_size=None,
+ log_sql_errors=True,
+ ):
+ results = []
+ truncated = False
+ description = ()
+
+ parsed_sql = parse_sql(sql, params)
+
+ table = self.conn.h5file.get_node(parsed_sql['from'])
+ table_rows = []
+ fields = parsed_sql['select']
+ colnames = ['value']
+ if type(table) is tables.table.Table:
+ colnames = table.colnames
+
+ query = ''
+ start = 0
+ end = table.nrows
+
+ # Use 'where' statement or get all the rows
+ def _cast_param(field, pname):
# Cast value to the column type
- coltype = table.coltypes[field]
+ coltype = table.dtype.name
+ if type(table) is tables.table.Table:
+ coltype = table.coltypes[field]
+ fcast = None
if coltype == 'string':
- params[value] = str(params[value])
+ fcast = str
elif coltype.startswith('int'):
- params[value] = int(params[value])
+ fcast = int
elif coltype.startswith('float'):
- params[value] = float(params[value])
+ fcast = float
+ if fcast:
+ params[pname] = fcast(params[pname])
+
+ def _translate_where(where):
+ # Translate SQL to PyTables expression
+ nonlocal start, end
+ expr = ''
+ operator = list(where)[0]
+
+ if operator in ['and', 'or']:
+ subexpr = [_translate_where(e) for e in where[operator]]
+ subexpr = filter(lambda e: e, subexpr)
+ subexpr = ["({})".format(e) for e in subexpr]
+ expr = " {} ".format(self.operators[operator]).join(subexpr)
+ elif operator == 'exists':
+ pass
+ elif where == {'eq': ['rowid', 'p0']}:
+ start = int(params['p0'])
+ end = start + 1
+ elif where == {'gt': ['rowid', 'p0']}:
+ start = int(params['p0']) + 1
+ else:
+ left, right = where[operator]
- translated = "{left} {operator} {right}".format(left=field, operator=operator, right=value)
- return translated, params
+ if isinstance(left, dict):
+ left = "(" + _translate_where(left) + ")"
+ elif left in params:
+ _cast_param(right, left)
-class Connection:
- def __init__(self, path):
- self.path = path
- self.h5file = tables.open_file(path)
+ if isinstance(right, dict):
+ right = "(" + _translate_where(right) + ")"
+ elif right in params:
+ _cast_param(left, right)
- def execute(self, sql, params=None, truncate=False):
- if params is None:
- params = {}
- rows = []
- truncated = False
- description = []
+ expr = "{left} {operator} {right}".format(
+ left=left,
+ operator=self.operators.get(operator, operator),
+ right=right,
+ )
- parsed_sql = _parse_sql(sql)
- table = self.h5file.get_node(parsed_sql['from'][1:-1])
- table_rows = []
- fields = parsed_sql['select'].split(',')
+ return expr
- # Use 'where' statement or get all the rows
if 'where' in parsed_sql:
- query = []
- start = 0
- end = table.nrows
- for condition in parsed_sql['where'].get_sublists():
- if str(condition) == '"rowid"=:p0':
- start = int(params['p0'])
- end = start + 1
- else:
- translated, params = _translate_condition(table, condition, params)
- query.append(translated)
- if query:
- query = ') & ('.join(query)
- query = '(' + query + ')'
- table_rows = table.where(query, params, start, end)
+ if type(parsed_sql['where']) is dict:
+ query = _translate_where(parsed_sql['where'])
else:
- table_rows = table.iterrows(start, end)
+ query = parsed_sql['where']
+
+ # Sort by column
+ orderby = ''
+ if 'orderby' in parsed_sql:
+ orderby = parsed_sql['orderby']
+ if type(orderby) is list:
+ orderby = orderby[0]
+ orderby = orderby['value']
+ if orderby == 'rowid':
+ orderby = ''
+
+ # Limit number of rows
+ limit = None
+ if 'limit' in parsed_sql:
+ limit = int(parsed_sql['limit'])
+
+ # Truncate if needed
+ if page_size and max_returned_rows and truncate:
+ if max_returned_rows == page_size:
+ max_returned_rows += 1
+
+ # Execute query
+ if query:
+ table_rows = table.where(query, params, start, end)
+ elif orderby:
+ table_rows = table.itersorted(orderby, start=start, stop=end)
else:
- table_rows = table.iterrows()
+ table_rows = table.iterrows(start, end)
# Prepare rows
- if len(fields) == 1 and fields[0] == 'count(*)':
- rows.append(Row({fields[0]: table.nrows}))
+ def normalize_field_value(value):
+ if type(value) is bytes:
+ return value.decode('utf-8')
+ elif not type(value) in (int, float, complex):
+ return str(value)
+ else:
+ return value
+
+ def make_get_rowid():
+ if type(table) is tables.table.Table:
+ def get_rowid(row):
+ return int(row.nrow)
+ else:
+ rowid = start - 1
+ def get_rowid(row):
+ nonlocal rowid
+ rowid += 1
+ return rowid
+ return get_rowid
+
+ def make_get_row_value():
+ if type(table) is tables.table.Table:
+ def get_row_value(row, field):
+ return row[field]
+ else:
+ def get_row_value(row, field):
+ return row
+ return get_row_value
+
+ if len(fields) == 1 and type(fields[0]['value']) is dict and \
+ fields[0]['value'].get('count') == '*':
+ results.append({'count(*)': int(table.nrows)})
else:
+ get_rowid = make_get_rowid()
+ get_row_value = make_get_row_value()
+ count = 0
for table_row in table_rows:
- row = Row()
+ count += 1
+ if limit and count > limit:
+ break
+ if truncate and max_returned_rows and count > max_returned_rows:
+ truncated = True
+ break
+ row = {}
for field in fields:
- if field == 'rowid':
- row[field] = table_row.nrow
- elif field == '*':
- for col in table.colnames:
- value = table_row[col]
- if type(value) is bytes:
- value = value.decode('utf-8')
- row[col] = value
+ field_name = field['value']
+ if type(field_name) is dict and 'distinct' in field_name:
+ field_name = field_name['distinct']
+ if field_name == 'rowid':
+ row['rowid'] = get_rowid(table_row)
+ elif field_name == '*':
+ for col in colnames:
+ row[col] = normalize_field_value(get_row_value(table_row, col))
else:
- row[field] = table_row[field]
- rows.append(row)
+ row[field_name] = normalize_field_value(get_row_value(table_row, field_name))
+ results.append(row)
# Prepare query description
- for field in fields:
+ for field in [f['value'] for f in fields]:
if field == '*':
- for col in table.colnames:
- description.append((col,))
+ for col in colnames:
+ description += ((col,),)
else:
- description.append((field,))
-
- # Return the rows
- if truncate:
- return rows, truncated, tuple(description)
- else:
- return rows
-
-class Row(OrderedDict):
- def __getitem__(self, label):
- if type(label) is int:
- return super(OrderedDict, self).__getitem__(list(self.keys())[label])
- else:
- return super(OrderedDict, self).__getitem__(label)
+ description += ((field,),)
- def __iter__(self):
- return self.values().__iter__()
+ return results, truncated, description