X-Git-Url: https://git.jsancho.org/?p=datasette-pytables.git;a=blobdiff_plain;f=datasette_pytables%2F__init__.py;h=93f9b89cfa0577b9d4b10215cec6b191c595c3d3;hp=0bbb25f9cd63eb55c4ab37212984d0aaea1a221a;hb=f2e02a0f692b0312c7e9e1051ff0cca6205b4972;hpb=ab2cd56eb3c5b15410fecdc19e037ff3b6ff6a63 diff --git a/datasette_pytables/__init__.py b/datasette_pytables/__init__.py index 0bbb25f..93f9b89 100644 --- a/datasette_pytables/__init__.py +++ b/datasette_pytables/__init__.py @@ -1,5 +1,5 @@ -from collections import OrderedDict -import sqlparse +from moz_sql_parser import parse +import re import tables _connector_type = 'pytables' @@ -11,7 +11,7 @@ def inspect(path): h5file = tables.open_file(path) for table in filter(lambda node: not(isinstance(node, tables.group.Group)), h5file): - colnames = [] + colnames = ['value'] if isinstance(table, tables.table.Table): colnames = table.colnames @@ -29,134 +29,252 @@ def inspect(path): 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 +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) -_operators = { - '=': '==', -} + try: + parsed = parse(sql) + except: + # Propably it's a PyTables expression + for token in ['group by', 'order by', 'limit', '']: + res = re.search('(?i)where (.*)' + token, sql) + if res: + modified_sql = re.sub('(?i)where (.*)(' + token + ')', '\g<2>', sql) + parsed = parse(modified_sql) + parsed['where'] = res.group(1).strip() + break -def _translate_condition(table, condition, params): - field = condition.left.get_real_name() + # Always a list of fields + if type(parsed['select']) is not list: + parsed['select'] = [parsed['select']] - 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 + return parsed - value = condition.right.value - if value.startswith(':'): - # Value is a parameters - value = value[1:] - if value in params: - # Cast value to the column type - coltype = table.coltypes[field] - if coltype == 'string': - params[value] = str(params[value]) - elif coltype.startswith('int'): - params[value] = int(params[value]) - elif coltype.startswith('float'): - params[value] = float(params[value]) - - translated = "{left} {operator} {right}".format(left=field, operator=operator, right=value) - return translated, params +_operators = { + 'eq': '==', + 'neq': '!=', + 'gt': '>', + 'gte': '>=', + 'lt': '<', + 'lte': '<=', + 'and': '&', + 'or': '|', +} class Connection: def __init__(self, path): self.path = path self.h5file = tables.open_file(path) - def execute(self, sql, params=None, truncate=False): + def execute(self, sql, params=None, truncate=False, page_size=None, max_returned_rows=None): if params is None: params = {} rows = [] truncated = False description = [] - parsed_sql = _parse_sql(sql) - table = self.h5file.get_node(parsed_sql['from'][1:-1]) + parsed_sql = _parse_sql(sql, params) + + if parsed_sql['from'] == 'sqlite_master': + rows = self._execute_datasette_query(sql, params) + description = (('value',)) + return rows, truncated, description + + table = self.h5file.get_node(parsed_sql['from']) table_rows = [] - fields = parsed_sql['select'].split(',') + 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.dtype.name + if type(table) is tables.table.Table: + coltype = table.coltypes[field] + fcast = None + if coltype == 'string': + fcast = str + elif coltype.startswith('int'): + fcast = int + elif coltype.startswith('float'): + 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(_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] + if left in params: + _cast_param(right, left) + elif right in params: + _cast_param(left, right) + + expr = "{left} {operator} {right}".format(left=left, operator=_operators.get(operator, operator), right=right) + + return expr + 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'] + + # 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) 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]: int(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') == '*': + rows.append(Row({'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: + count += 1 + if limit and count > limit: + break + if truncate and max_returned_rows and count > max_returned_rows: + truncated = True + break row = Row() for field in fields: - if field == 'rowid': - row[field] = int(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] + row[field_name] = normalize_field_value(get_row_value(table_row, field_name)) rows.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: + for col in colnames: description.append((col,)) else: description.append((field,)) # Return the rows - if truncate: - return rows, truncated, tuple(description) + return rows, truncated, tuple(description) + + def _execute_datasette_query(self, sql, params): + "Datasette special queries for getting tables info" + if sql == "SELECT count(*) from sqlite_master WHERE type = 'view' and name=:n": + row = Row() + row['count(*)'] = 0 + return [row] + elif sql == 'select sql from sqlite_master where name = :n and type="table"': + try: + table = self.h5file.get_node(params['n']) + colnames = ['value'] + if type(table) is tables.table.Table: + colnames = table.colnames + row = Row() + row['sql'] = 'CREATE TABLE {} ({})'.format(params['n'], ", ".join(colnames)) + return [row] + except: + return [] + else: + raise Exception("SQLite queries cannot be executed with this connector") + + +class Row(list): + def __init__(self, values=None): + self.labels = [] + self.values = [] + if values: + for idx in values: + self.__setitem__(idx, values[idx]) + + def __setitem__(self, idx, value): + if type(idx) is str: + if idx in self.labels: + self.values[self.labels.index(idx)] = value + else: + self.labels.append(idx) + self.values.append(value) else: - return rows + self.values[idx] = value -class Row(OrderedDict): - def __getitem__(self, label): - if type(label) is int: - return super(OrderedDict, self).__getitem__(list(self.keys())[label]) + def __getitem__(self, idx): + if type(idx) is str: + return self.values[self.labels.index(idx)] else: - return super(OrderedDict, self).__getitem__(label) + return self.values[idx] def __iter__(self): - return self.values().__iter__() + return self.values.__iter__()