X-Git-Url: https://git.jsancho.org/?p=datasette-pytables.git;a=blobdiff_plain;f=datasette_pytables%2F__init__.py;h=e8d0a6ccb4519f72d2e0680e0c3b4b8f44dc2574;hp=d44c9f93485e4e187c4ec06f3565da0df6723684;hb=2cf7538022a4290be1cdbe6f8cf13e44f8190bee;hpb=8608557baec40b36d79ab47379fe735b69e0f2fa

diff --git a/datasette_pytables/__init__.py b/datasette_pytables/__init__.py
index d44c9f9..e8d0a6c 100644
--- a/datasette_pytables/__init__.py
+++ b/datasette_pytables/__init__.py
@@ -1,6 +1,7 @@
-import tables
-import sqlparse
 from collections import OrderedDict
+from moz_sql_parser import parse
+import re
+import tables
 
 _connector_type = 'pytables'
 
@@ -19,7 +20,7 @@ def inspect(path):
             'name': table._v_pathname,
             'columns': colnames,
             'primary_keys': [],
-            'count': table.nrows,
+            'count': int(table.nrows),
             'label_column': None,
             'hidden': False,
             'fts_table': None,
@@ -29,24 +30,29 @@ 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] = ""
-        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)
+
+    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):
@@ -54,45 +60,106 @@ class Connection:
         self.h5file = tables.open_file(path)
 
     def execute(self, sql, params=None, truncate=False):
+        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)
+        table = self.h5file.get_node(parsed_sql['from'])
         table_rows = []
-        fields = parsed_sql['select'].split(',')
+        fields = parsed_sql['select']
+
+        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.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
+            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)
+            elif where == {'eq': ['rowid', 'p0']}:
+                nonlocal start, end
+                start = int(params['p0'])
+                end = start + 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:
-            pass
+            try:
+                query = _translate_where(parsed_sql['where'])
+            except:
+                # Probably it's a PyTables query
+                query = str(parsed_sql['where'])[6:]    # without where keyword
+
+        # Limit number of rows
+        if 'limit' in parsed_sql:
+            max_rows = int(parsed_sql['limit'])
+            if end - start > max_rows:
+                end = start + max_rows
+
+        # 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]: table.nrows}))
+        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 == 'rowid':
-                        row[field] = table_row.nrow
-                    elif field == '*':
+                    if field['value'] == 'rowid':
+                        row['rowid'] = int(table_row.nrow)
+                    elif field['value'] == '*':
                         for col in table.colnames:
-                            row[col] = table_row[col]
+                            value = table_row[col]
+                            if type(value) is bytes:
+                                value = value.decode('utf-8')
+                            row[col] = value
                     else:
-                        row[field] = table_row[field]
+                        row[field['value']] = table_row[field['value']]
                 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:
                     description.append((col,))
             else:
                 description.append((field,))
 
+        # Return the rows
         if truncate:
             return rows, truncated, tuple(description)
         else:
@@ -102,6 +169,8 @@ 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)
 
     def __iter__(self):
         return self.values().__iter__()