datasette_pytables/__init__.py

   1 import tables
   2 import datasette_connectors as dc
   3 from .utils import parse_sql
   4
   5
   6 class PyTablesConnection(dc.Connection):
   7     def __init__(self, path, connector):
   8         super().__init__(path, connector)
   9         self.h5file = tables.open_file(path)
  10
  11
  12 class PyTablesConnector(dc.Connector):
  13     connector_type = 'pytables'
  14     connection_class = PyTablesConnection
  15
  16     operators = {
  17         'eq': '==',
  18         'neq': '!=',
  19         'gt': '>',
  20         'gte': '>=',
  21         'lt': '<',
  22         'lte': '<=',
  23         'and': '&',
  24         'or': '|',
  25         'binary_and': '&',
  26         'binary_or': '|',
  27     }
  28
  29     def table_names(self):
  30         return [
  31             node._v_pathname
  32             for node in self.conn.h5file
  33             if not(isinstance(node, tables.group.Group))
  34         ]
  35
  36     def table_count(self, table_name):
  37         table = self.conn.h5file.get_node(table_name)
  38         return int(table.nrows)
  39
  40     def table_info(self, table_name):
  41         table = self.conn.h5file.get_node(table_name)
  42         colnames = ['value']
  43         if isinstance(table, tables.table.Table):
  44             colnames = table.colnames
  45
  46         return [
  47             {
  48                 'idx': idx,
  49                 'name': colname,
  50                 'primary_key': False,
  51             }
  52             for idx, colname in enumerate(colnames)
  53         ]
  54
  55     def hidden_table_names(self):
  56         return []
  57
  58     def detect_spatialite(self):
  59         return False
  60
  61     def view_names(self):
  62         return []
  63
  64     def detect_fts(self, table_name):
  65         return False
  66
  67     def foreign_keys(self, table_name):
  68         return []
  69
  70     def table_exists(self, table_name):
  71         try:
  72             self.conn.h5file.get_node(table_name)
  73             return True
  74         except:
  75             return False
  76
  77     def table_definition(self, table_type, table_name):
  78         table = self.conn.h5file.get_node(table_name)
  79         colnames = ['value']
  80         if isinstance(table, tables.table.Table):
  81             colnames = table.colnames
  82
  83         return 'CREATE TABLE {} ({})'.format(
  84             table_name,
  85             ', '.join(colnames),
  86         )
  87
  88     def indices_definition(self, table_name):
  89         return []
  90
  91     def execute(
  92         self,
  93         sql,
  94         params=None,
  95         truncate=False,
  96         custom_time_limit=None,
  97         page_size=None,
  98         log_sql_errors=True,
  99     ):
 100         results = []
 101         truncated = False
 102         description = ()
 103
 104         parsed_sql = parse_sql(sql, params)
 105
 106         while isinstance(parsed_sql['from'], dict):
 107             # Pytables does not support subqueries
 108             parsed_sql['from'] = parsed_sql['from']['value']['from']
 109
 110         table = self.conn.h5file.get_node(parsed_sql['from'])
 111         table_rows = []
 112         fields = parsed_sql['select']
 113         colnames = ['value']
 114         if type(table) is tables.table.Table:
 115             colnames = table.colnames
 116
 117         query = ''
 118         start = 0
 119         end = table.nrows
 120
 121         # Use 'where' statement or get all the rows
 122         def _cast_param(field, pname):
 123             # Cast value to the column type
 124             coltype = table.dtype.name
 125             if type(table) is tables.table.Table:
 126                 coltype = table.coltypes[field]
 127             fcast = None
 128             if coltype == 'string':
 129                 fcast = str
 130             elif coltype.startswith('int'):
 131                 fcast = int
 132             elif coltype.startswith('float'):
 133                 fcast = float
 134             if fcast:
 135                 params[pname] = fcast(params[pname])
 136
 137         def _translate_where(where):
 138             # Translate SQL to PyTables expression
 139             nonlocal start, end
 140             expr = ''
 141             operator = list(where)[0]
 142
 143             if operator in ['and', 'or']:
 144                 subexpr = [_translate_where(e) for e in where[operator]]
 145                 subexpr = filter(lambda e: e, subexpr)
 146                 subexpr = ["({})".format(e) for e in subexpr]
 147                 expr = " {} ".format(self.operators[operator]).join(subexpr)
 148             elif operator == 'exists':
 149                 pass
 150             elif where == {'eq': ['rowid', 'p0']}:
 151                 start = int(params['p0'])
 152                 end = start + 1
 153             elif where == {'gt': ['rowid', 'p0']}:
 154                 start = int(params['p0']) + 1
 155             else:
 156                 left, right = where[operator]
 157
 158                 if isinstance(left, dict):
 159                     left = "(" + _translate_where(left) + ")"
 160                 elif left in params:
 161                     _cast_param(right, left)
 162
 163                 if isinstance(right, dict):
 164                     right = "(" + _translate_where(right) + ")"
 165                 elif right in params:
 166                     _cast_param(left, right)
 167
 168                 expr = "{left} {operator} {right}".format(
 169                     left=left,
 170                     operator=self.operators.get(operator, operator),
 171                     right=right,
 172                 )
 173
 174             return expr
 175
 176         if 'where' in parsed_sql:
 177             if type(parsed_sql['where']) is dict:
 178                 query = _translate_where(parsed_sql['where'])
 179             else:
 180                 query = parsed_sql['where']
 181
 182         # Sort by column
 183         orderby = ''
 184         if 'orderby' in parsed_sql:
 185             orderby = parsed_sql['orderby']
 186             if type(orderby) is list:
 187                 orderby = orderby[0]
 188             orderby = orderby['value']
 189             if orderby == 'rowid':
 190                 orderby = ''
 191
 192         # Limit number of rows
 193         limit = None
 194         if 'limit' in parsed_sql:
 195             limit = int(parsed_sql['limit'])
 196
 197         # Truncate if needed
 198         if page_size and max_returned_rows and truncate:
 199             if max_returned_rows == page_size:
 200                 max_returned_rows += 1
 201
 202         # Execute query
 203         if query:
 204             if not ' glob ' in query:
 205                 table_rows = table.where(query, params, start, end)
 206         elif orderby:
 207             table_rows = table.itersorted(orderby, start=start, stop=end)
 208         else:
 209             table_rows = table.iterrows(start, end)
 210
 211         # Prepare rows
 212         def normalize_field_value(value):
 213             if type(value) is bytes:
 214                 return value.decode('utf-8')
 215             elif not type(value) in (int, float, complex):
 216                 return str(value)
 217             else:
 218                 return value
 219
 220         def make_get_rowid():
 221             if type(table) is tables.table.Table:
 222                 def get_rowid(row):
 223                     return int(row.nrow)
 224             else:
 225                 rowid = start - 1
 226                 def get_rowid(row):
 227                     nonlocal rowid
 228                     rowid += 1
 229                     return rowid
 230             return get_rowid
 231
 232         def make_get_row_value():
 233             if type(table) is tables.table.Table:
 234                 def get_row_value(row, field):
 235                     return row[field]
 236             else:
 237                 def get_row_value(row, field):
 238                     return row
 239             return get_row_value
 240
 241         # Get results
 242         get_rowid = make_get_rowid()
 243         get_row_value = make_get_row_value()
 244         count = 0
 245         for table_row in table_rows:
 246             count += 1
 247             if limit is not None and count > limit:
 248                 break
 249             if truncate and max_returned_rows and count > max_returned_rows:
 250                 truncated = True
 251                 break
 252             row = {}
 253             for field in fields:
 254                 field_name = field
 255                 if isinstance(field, dict):
 256                     field_name = field['value']
 257                 if isinstance(field_name, dict) and 'distinct' in field_name:
 258                     field_name = field_name['distinct']
 259                 if field_name == 'rowid':
 260                     row['rowid'] = get_rowid(table_row)
 261                 elif field_name == '*':
 262                     for col in colnames:
 263                         row[col] = normalize_field_value(get_row_value(table_row, col))
 264                 elif isinstance(field_name, dict):
 265                     if field_name.get('count') == '*':
 266                         row['count(*)'] = int(table.nrows)
 267                     elif field_name.get('json_type'):
 268                         field_name = field_name.get('json_type')
 269                         row['json_type(' + field_name + ')'] = table.coltypes[field_name]
 270                     else:
 271                         raise Exception("Function not recognized")
 272                 else:
 273                     row[field_name] = normalize_field_value(get_row_value(table_row, field_name))
 274             results.append(row)
 275
 276         # Prepare query description
 277         for field in [f['value'] if isinstance(f, dict) else f for f in fields]:
 278             if field == '*':
 279                 for col in colnames:
 280                     description += ((col,),)
 281             else:
 282                 description += ((field,),)
 283
 284         return results, truncated, description