datasette_pytables/__init__.py

   1 from collections import OrderedDict
   2 from moz_sql_parser import parse
   3 import re
   4 import tables
   5
   6 _connector_type = 'pytables'
   7
   8 def inspect(path):
   9     "Open file and return tables info"
  10     h5tables = {}
  11     views = []
  12     h5file = tables.open_file(path)
  13
  14     for table in filter(lambda node: not(isinstance(node, tables.group.Group)), h5file):
  15         colnames = []
  16         if isinstance(table, tables.table.Table):
  17             colnames = table.colnames
  18
  19         h5tables[table._v_pathname] = {
  20             'name': table._v_pathname,
  21             'columns': colnames,
  22             'primary_keys': [],
  23             'count': int(table.nrows),
  24             'label_column': None,
  25             'hidden': False,
  26             'fts_table': None,
  27             'foreign_keys': {'incoming': [], 'outgoing': []},
  28         }
  29
  30     h5file.close()
  31     return h5tables, views, _connector_type
  32
  33 def _parse_sql(sql, params):
  34     # Table name
  35     sql = re.sub('(?i)from \[(.*)]', 'from "\g<1>"', sql)
  36     # Params
  37     for param in params:
  38         sql = sql.replace(":" + param, param)
  39
  40     try:
  41         parsed = parse(sql)
  42     except:
  43         # Propably it's a PyTables expression
  44         for token in ['group by', 'order by', 'limit']:
  45             res = re.search('(?i)where (.*)' + token, sql)
  46             if res:
  47                 modified_sql = re.sub('(?i)where (.*)(' + token + ')', '\g<2>', sql)
  48                 parsed = parse(modified_sql)
  49                 parsed['where'] = res.group(1)
  50                 break
  51
  52     # Always a list of fields
  53     if type(parsed['select']) is not list:
  54         parsed['select'] = [parsed['select']]
  55
  56     return parsed
  57
  58 _operators = {
  59     'eq': '==',
  60     'neq': '!=',
  61     'gt': '>',
  62     'gte': '>=',
  63     'lt': '<',
  64     'lte': '<=',
  65     'and': '&',
  66     'or': '|',
  67 }
  68
  69 class Connection:
  70     def __init__(self, path):
  71         self.path = path
  72         self.h5file = tables.open_file(path)
  73
  74     def execute(self, sql, params=None, truncate=False):
  75         if params is None:
  76             params = {}
  77         rows = []
  78         truncated = False
  79         description = []
  80
  81         parsed_sql = _parse_sql(sql, params)
  82         table = self.h5file.get_node(parsed_sql['from'])
  83         table_rows = []
  84         fields = parsed_sql['select']
  85
  86         query = ''
  87         start = 0
  88         end = table.nrows
  89
  90         # Use 'where' statement or get all the rows
  91         def _cast_param(field, pname):
  92             # Cast value to the column type
  93             coltype = table.coltypes[field]
  94             fcast = None
  95             if coltype == 'string':
  96                 fcast = str
  97             elif coltype.startswith('int'):
  98                 fcast = int
  99             elif coltype.startswith('float'):
 100                 fcast = float
 101             if fcast:
 102                 params[pname] = fcast(params[pname])
 103
 104         def _translate_where(where):
 105             # Translate SQL to PyTables expression
 106             expr = ''
 107             operator = list(where)[0]
 108
 109             if operator in ['and', 'or']:
 110                 subexpr = ["({})".format(_translate_where(q)) for q in where[operator]]
 111                 expr = " {} ".format(_operators[operator]).join(subexpr)
 112             elif where == {'eq': ['rowid', 'p0']}:
 113                 nonlocal start, end
 114                 start = int(params['p0'])
 115                 end = start + 1
 116             else:
 117                 left, right = where[operator]
 118                 if left in params:
 119                     _cast_param(right, left)
 120                 elif right in params:
 121                     _cast_param(left, right)
 122
 123                 expr = "{left} {operator} {right}".format(left=left, operator=_operators.get(operator, operator), right=right)
 124
 125             return expr
 126
 127         if 'where' in parsed_sql:
 128             if type(parsed_sql['where']) is dict:
 129                 query = _translate_where(parsed_sql['where'])
 130             else:
 131                 query = parsed_sql['where']
 132
 133         # Limit number of rows
 134         if 'limit' in parsed_sql:
 135             max_rows = int(parsed_sql['limit'])
 136             if end - start > max_rows:
 137                 end = start + max_rows
 138
 139         # Execute query
 140         if query:
 141             table_rows = table.where(query, params, start, end)
 142         else:
 143             table_rows = table.iterrows(start, end)
 144
 145         # Prepare rows
 146         if len(fields) == 1 and type(fields[0]['value']) is dict and \
 147            fields[0]['value'].get('count') == '*':
 148             rows.append(Row({'count(*)': int(table.nrows)}))
 149         else:
 150             for table_row in table_rows:
 151                 row = Row()
 152                 for field in fields:
 153                     if field['value'] == 'rowid':
 154                         row['rowid'] = int(table_row.nrow)
 155                     elif field['value'] == '*':
 156                         for col in table.colnames:
 157                             value = table_row[col]
 158                             if type(value) is bytes:
 159                                 value = value.decode('utf-8')
 160                             row[col] = value
 161                     else:
 162                         row[field['value']] = table_row[field['value']]
 163                 rows.append(row)
 164
 165         # Prepare query description
 166         for field in [f['value'] for f in fields]:
 167             if field == '*':
 168                 for col in table.colnames:
 169                     description.append((col,))
 170             else:
 171                 description.append((field,))
 172
 173         # Return the rows
 174         if truncate:
 175             return rows, truncated, tuple(description)
 176         else:
 177             return rows
 178
 179 class Row(OrderedDict):
 180     def __getitem__(self, label):
 181         if type(label) is int:
 182             return super(OrderedDict, self).__getitem__(list(self.keys())[label])
 183         else:
 184             return super(OrderedDict, self).__getitem__(label)
 185
 186     def __iter__(self):
 187         return self.values().__iter__()