datasette_pytables/__init__.py

   1 from moz_sql_parser import parse
   2 import re
   3 import tables
   4
   5 _connector_type = 'pytables'
   6
   7 def inspect(path):
   8     "Open file and return tables info"
   9     h5tables = {}
  10     views = []
  11     h5file = tables.open_file(path)
  12
  13     for table in filter(lambda node: not(isinstance(node, tables.group.Group)), h5file):
  14         colnames = []
  15         if isinstance(table, tables.table.Table):
  16             colnames = table.colnames
  17
  18         h5tables[table._v_pathname] = {
  19             'name': table._v_pathname,
  20             'columns': colnames,
  21             'primary_keys': [],
  22             'count': int(table.nrows),
  23             'label_column': None,
  24             'hidden': False,
  25             'fts_table': None,
  26             'foreign_keys': {'incoming': [], 'outgoing': []},
  27         }
  28
  29     h5file.close()
  30     return h5tables, views, _connector_type
  31
  32 def _parse_sql(sql, params):
  33     # Table name
  34     sql = re.sub('(?i)from \[(.*)]', 'from "\g<1>"', sql)
  35     # Params
  36     for param in params:
  37         sql = sql.replace(":" + param, param)
  38
  39     try:
  40         parsed = parse(sql)
  41     except:
  42         # Propably it's a PyTables expression
  43         for token in ['group by', 'order by', 'limit', '']:
  44             res = re.search('(?i)where (.*)' + token, sql)
  45             if res:
  46                 modified_sql = re.sub('(?i)where (.*)(' + token + ')', '\g<2>', sql)
  47                 parsed = parse(modified_sql)
  48                 parsed['where'] = res.group(1).strip()
  49                 break
  50
  51     # Always a list of fields
  52     if type(parsed['select']) is not list:
  53         parsed['select'] = [parsed['select']]
  54
  55     return parsed
  56
  57 _operators = {
  58     'eq': '==',
  59     'neq': '!=',
  60     'gt': '>',
  61     'gte': '>=',
  62     'lt': '<',
  63     'lte': '<=',
  64     'and': '&',
  65     'or': '|',
  66 }
  67
  68 class Connection:
  69     def __init__(self, path):
  70         self.path = path
  71         self.h5file = tables.open_file(path)
  72
  73     def execute(self, sql, params=None, truncate=False, page_size=None, max_returned_rows=None):
  74         if params is None:
  75             params = {}
  76         rows = []
  77         truncated = False
  78         description = []
  79
  80         parsed_sql = _parse_sql(sql, params)
  81
  82         if parsed_sql['from'] == 'sqlite_master':
  83             rows = self._execute_datasette_query(sql, params)
  84             description = (('value',))
  85             return rows, truncated, description
  86
  87         table = self.h5file.get_node(parsed_sql['from'])
  88         table_rows = []
  89         fields = parsed_sql['select']
  90
  91         query = ''
  92         start = 0
  93         end = table.nrows
  94
  95         # Use 'where' statement or get all the rows
  96         def _cast_param(field, pname):
  97             # Cast value to the column type
  98             if type(table) is tables.table.Table:
  99                 coltype = table.coltypes[field]
 100             else:
 101                 coltype = table.dtype.name
 102             fcast = None
 103             if coltype == 'string':
 104                 fcast = str
 105             elif coltype.startswith('int'):
 106                 fcast = int
 107             elif coltype.startswith('float'):
 108                 fcast = float
 109             if fcast:
 110                 params[pname] = fcast(params[pname])
 111
 112         def _translate_where(where):
 113             # Translate SQL to PyTables expression
 114             nonlocal start, end
 115             expr = ''
 116             operator = list(where)[0]
 117
 118             if operator in ['and', 'or']:
 119                 subexpr = [_translate_where(e) for e in where[operator]]
 120                 subexpr = filter(lambda e: e, subexpr)
 121                 subexpr = ["({})".format(e) for e in subexpr]
 122                 expr = " {} ".format(_operators[operator]).join(subexpr)
 123             elif operator == 'exists':
 124                 pass
 125             elif where == {'eq': ['rowid', 'p0']}:
 126                 start = int(params['p0'])
 127                 end = start + 1
 128             elif where == {'gt': ['rowid', 'p0']}:
 129                 start = int(params['p0']) + 1
 130             else:
 131                 left, right = where[operator]
 132                 if left in params:
 133                     _cast_param(right, left)
 134                 elif right in params:
 135                     _cast_param(left, right)
 136
 137                 expr = "{left} {operator} {right}".format(left=left, operator=_operators.get(operator, operator), right=right)
 138
 139             return expr
 140
 141         if 'where' in parsed_sql:
 142             if type(parsed_sql['where']) is dict:
 143                 query = _translate_where(parsed_sql['where'])
 144             else:
 145                 query = parsed_sql['where']
 146
 147         # Limit number of rows
 148         limit = None
 149         if 'limit' in parsed_sql:
 150             limit = int(parsed_sql['limit'])
 151
 152         # Truncate if needed
 153         if page_size and max_returned_rows and truncate:
 154             if max_returned_rows == page_size:
 155                 max_returned_rows += 1
 156
 157         # Execute query
 158         if query:
 159             table_rows = table.where(query, params, start, end)
 160         else:
 161             table_rows = table.iterrows(start, end)
 162
 163         # Prepare rows
 164         if len(fields) == 1 and type(fields[0]['value']) is dict and \
 165            fields[0]['value'].get('count') == '*':
 166             rows.append(Row({'count(*)': int(table.nrows)}))
 167         else:
 168             if type(table) is tables.table.Table:
 169                 count = 0
 170                 for table_row in table_rows:
 171                     count += 1
 172                     if limit and count > limit:
 173                         break
 174                     if truncate and max_returned_rows and count > max_returned_rows:
 175                         truncated = True
 176                         break
 177                     row = Row()
 178                     for field in fields:
 179                         field_name = field['value']
 180                         if type(field_name) is dict and 'distinct' in field_name:
 181                             field_name = field_name['distinct']
 182                         if field_name == 'rowid':
 183                             row['rowid'] = int(table_row.nrow)
 184                         elif field_name == '*':
 185                             for col in table.colnames:
 186                                 value = table_row[col]
 187                                 if type(value) is bytes:
 188                                     value = value.decode('utf-8')
 189                                 row[col] = value
 190                         else:
 191                             row[field_name] = table_row[field_name]
 192                     rows.append(row)
 193             else:
 194                 # Any kind of array
 195                 rowid = start - 1
 196                 count = 0
 197                 for table_row in table_rows:
 198                     count += 1
 199                     if limit and count > limit:
 200                         break
 201                     if truncate and max_returned_rows and count > max_returned_rows:
 202                         truncated = True
 203                         break
 204                     row = Row()
 205                     rowid += 1
 206                     for field in fields:
 207                         field_name = field['value']
 208                         if type(field_name) is dict and 'distinct' in field_name:
 209                             field_name = field_name['distinct']
 210                         if field_name == 'rowid':
 211                             row['rowid'] = rowid
 212                         else:
 213                             value = table_row
 214                             if type(value) is bytes:
 215                                 value = value.decode('utf-8')
 216                             row['value'] = value
 217                     rows.append(row)
 218
 219         # Prepare query description
 220         for field in [f['value'] for f in fields]:
 221             if field == '*':
 222                 if type(table) is tables.table.Table:
 223                     for col in table.colnames:
 224                         description.append((col,))
 225                 else:
 226                     description.append(('value',))
 227             else:
 228                 description.append((field,))
 229
 230         # Return the rows
 231         return rows, truncated, tuple(description)
 232
 233     def _execute_datasette_query(self, sql, params):
 234         "Datasette special queries for getting tables info"
 235         if sql == "SELECT count(*) from sqlite_master WHERE type = 'view' and name=:n":
 236             row = Row()
 237             row['count(*)'] = 0
 238             return [row]
 239         elif sql == 'select sql from sqlite_master where name = :n and type="table"':
 240             try:
 241                 table = self.h5file.get_node(params['n'])
 242                 row = Row()
 243                 row['sql'] = 'CREATE TABLE {} ()'.format(params['n'])
 244                 return [row]
 245             except:
 246                 return []
 247         else:
 248             raise Exception("SQLite queries cannot be executed with this connector")
 249
 250 class Row(list):
 251     def __init__(self, values=None):
 252         self.labels = []
 253         self.values = []
 254         if values:
 255             for idx in values:
 256                 self.__setitem__(idx, values[idx])
 257
 258     def __setitem__(self, idx, value):
 259         if type(idx) is str:
 260             if idx in self.labels:
 261                 self.values[self.labels.index(idx)] = value
 262             else:
 263                 self.labels.append(idx)
 264                 self.values.append(value)
 265         else:
 266             self.values[idx] = value
 267
 268     def __getitem__(self, idx):
 269         if type(idx) is str:
 270             return self.values[self.labels.index(idx)]
 271         else:
 272             return self.values[idx]
 273
 274     def __iter__(self):
 275         return self.values.__iter__()