X-Git-Url: https://git.jsancho.org/?p=datasette-pytables.git;a=blobdiff_plain;f=datasette_pytables%2F__init__.py;h=4c1cc0f9837aff45375c1f6c1654b8e71d359094;hp=e8d0a6ccb4519f72d2e0680e0c3b4b8f44dc2574;hb=29e72871ce6657fd858dddb1689de8b64d7b6f54;hpb=2cf7538022a4290be1cdbe6f8cf13e44f8190bee diff --git a/datasette_pytables/__init__.py b/datasette_pytables/__init__.py index e8d0a6c..4c1cc0f 100644 --- a/datasette_pytables/__init__.py +++ b/datasette_pytables/__init__.py @@ -1,84 +1,158 @@ -from collections import OrderedDict -from moz_sql_parser import parse -import re 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) + + +class PyTablesConnector(dc.Connector): + connector_type = 'pytables' + connection_class = PyTablesConnection + + operators = { + 'eq': '==', + 'neq': '!=', + 'gt': '>', + 'gte': '>=', + 'lt': '<', + 'lte': '<=', + 'and': '&', + 'or': '|', + 'binary_and': '&', + 'binary_or': '|', + } + + def _serialize_table_name(self, table_name): + return table_name.replace('/', '%') + + def _deserialize_table_name(self, table_name): + return table_name.replace('%', '/') - for table in filter(lambda node: not(isinstance(node, tables.group.Group)), h5file): - colnames = [] + def table_names(self): + return [ + self._serialize_table_name(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(self._deserialize_table_name(table_name)) + return int(table.nrows) + + def table_info(self, table_name): + table = self.conn.h5file.get_node(self._deserialize_table_name(table_name)) + columns = [ + { + 'name': 'value', + 'type': table.dtype.name, + } + ] + if isinstance(table, tables.table.Table): + columns = [ + { + 'name': colname, + 'type': table.coltypes[colname], + } + for colname in table.colnames + ] + + return [ + { + 'cid': cid, + 'name': column['name'], + 'type': column['type'], + 'notnull': True, + 'default_value': None, + 'is_pk': False, + } + for cid, column in enumerate(columns) + ] + + 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 table_exists(self, table_name): + try: + self.conn.h5file.get_node(self._deserialize_table_name(table_name)) + return True + except: + return False + + def table_definition(self, table_type, table_name): + table_name = self._deserialize_table_name(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': int(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, params): - # Table name - sql = re.sub('(?i)from \[(.*)]', 'from "\g<1>"', sql) - # Params - for param in params: - sql = sql.replace(":" + param, param) - - parsed = parse(sql) - if type(parsed['select']) is not list: - parsed['select'] = [parsed['select']] - - return parsed - -_operators = { - 'eq': '==', - 'neq': '!=', - 'gt': '>', - 'gte': '>=', - 'lt': '<', - 'lte': '<=', - 'and': '&', - 'or': '|', -} - -class Connection: - def __init__(self, path): - self.path = path - self.h5file = tables.open_file(path) + return 'CREATE TABLE {} ({})'.format( + table_name, + ', '.join(colnames), + ) + + def indices_definition(self, table_name): + return [] - def execute(self, sql, params=None, truncate=False): - if params is None: - params = {} - rows = [] + def execute( + self, + sql, + params=None, + truncate=False, + custom_time_limit=None, + page_size=None, + log_sql_errors=True, + ): + results = [] truncated = False - description = [] + description = () + + # Some Datasette queries uses glob operand, not supported by Pytables + if ' glob ' in sql: + return results, truncated, description + + parsed_sql = parse_sql(sql, params) - parsed_sql = _parse_sql(sql, params) - table = self.h5file.get_node(parsed_sql['from']) + while isinstance(parsed_sql['from'], dict): + # Pytables does not support subqueries + parsed_sql['from'] = parsed_sql['from']['value']['from'] + + table = self.conn.h5file.get_node(self._deserialize_table_name(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 + def _get_field_type(field): + coltype = table.dtype.name + if type(table) is tables.table.Table: + coltype = table.coltypes[field] + return coltype + # Use 'where' statement or get all the rows def _cast_param(field, pname): # Cast value to the column type - coltype = table.coltypes[field] + coltype = _get_field_type(field) fcast = None if coltype == 'string': fcast = str @@ -91,86 +165,155 @@ class Connection: def _translate_where(where): # Translate SQL to PyTables expression + nonlocal start, end expr = '' operator = list(where)[0] if operator in ['and', 'or']: - subexpr = ["({})".format(_translate_where(q)) for q in where[operator]] - expr = " {} ".format(_operators[operator]).join(subexpr) + 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']}: - nonlocal start, end start = int(params['p0']) end = start + 1 + elif where == {'gt': ['rowid', 'p0']}: + start = int(params['p0']) + 1 else: left, right = where[operator] - if left in params: + + if isinstance(left, dict): + left = "(" + _translate_where(left) + ")" + elif left in params: _cast_param(right, left) + + if isinstance(right, dict): + right = "(" + _translate_where(right) + ")" elif right in params: _cast_param(left, right) - expr = "{left} {operator} {right}".format(left=left, operator=_operators.get(operator, operator), right=right) + expr = "{left} {operator} {right}".format( + left=left, + operator=self.operators.get(operator, operator), + right=right, + ) return expr if 'where' in parsed_sql: - try: + if type(parsed_sql['where']) is dict: query = _translate_where(parsed_sql['where']) - except: - # Probably it's a PyTables query - query = str(parsed_sql['where'])[6:] # without where keyword + else: + 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: - max_rows = int(parsed_sql['limit']) - if end - start > max_rows: - end = start + max_rows + limit = int(parsed_sql['limit']) + + # Offset + offset = None + if 'offset' in parsed_sql: + offset = int(parsed_sql['offset']) + + # 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(start, end) # Prepare rows - if len(fields) == 1 and type(fields[0]['value']) is dict and \ - fields[0]['value'].get('count') == '*': - rows.append(Row({'count(*)': int(table.nrows)})) - else: - for table_row in table_rows: - row = Row() - for field in fields: - if field['value'] == 'rowid': - row['rowid'] = int(table_row.nrow) - elif field['value'] == '*': - for col in table.colnames: - value = table_row[col] - if type(value) is bytes: - value = value.decode('utf-8') - row[col] = value + 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 + + # Get results + get_rowid = make_get_rowid() + get_row_value = make_get_row_value() + if offset: + table_rows = table_rows[offset:] + count = 0 + for table_row in table_rows: + count += 1 + if limit is not None and count > limit: + break + if truncate and max_returned_rows and count > max_returned_rows: + truncated = True + break + row = {} + for field in fields: + field_name = field + if isinstance(field, dict): + field_name = field['value'] + if isinstance(field_name, 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)) + elif isinstance(field_name, dict): + if field_name.get('count') == '*': + row['count(*)'] = int(table.nrows) + elif field_name.get('json_type'): + field_name = field_name.get('json_type') + row['json_type(' + field_name + ')'] = _get_field_type(field_name) else: - row[field['value']] = table_row[field['value']] - rows.append(row) + raise Exception("Function not recognized") + else: + row[field_name] = normalize_field_value(get_row_value(table_row, field_name)) + results.append(row) # Prepare query description - for field in [f['value'] for f in fields]: + for field in [f['value'] if isinstance(f, dict) else f 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