X-Git-Url: https://git.jsancho.org/?p=datasette-pytables.git;a=blobdiff_plain;f=datasette_pytables%2F__init__.py;h=217214f6f4d7ece9b9566e8018f85ee85c5ee47f;hp=8516315cd49bbf4bf4507590d3e47e5a4878a1b1;hb=1d4a2abc6556f51a94b28f1e06cd682be961a7ea;hpb=d9e3d4c6f66634a32236b238ce04cefff17f79b1

diff --git a/datasette_pytables/__init__.py b/datasette_pytables/__init__.py
index 8516315..217214f 100644
--- a/datasette_pytables/__init__.py
+++ b/datasette_pytables/__init__.py
@@ -1,5 +1,9 @@
+from collections import OrderedDict
+import sqlparse
 import tables
 
+_connector_type = 'pytables'
+
 def inspect(path):
     "Open file and return tables info"
     h5tables = {}
@@ -23,4 +27,136 @@ def inspect(path):
         }
 
     h5file.close()
-    return h5tables, views
+    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
+
+_operators = {
+    '=': '==',
+}
+
+def _translate_condition(table, condition, params):
+    field = condition.left.get_real_name()
+
+    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
+
+    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
+
+class Connection:
+    def __init__(self, path):
+        self.path = path
+        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])
+        table_rows = []
+        fields = parsed_sql['select'].split(',')
+
+        # Use 'where' statement or get all the rows
+        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)
+            else:
+                table_rows = table.iterrows(start, end)
+        else:
+            table_rows = table.iterrows()
+
+        # Prepare rows
+        if len(fields) == 1 and fields[0] == 'count(*)':
+            rows.append(Row({fields[0]: 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 == '*':
+                        for col in table.colnames:
+                            value = table_row[col]
+                            if type(value) is bytes:
+                                value = value.decode('utf-8')
+                            row[col] = value
+                    else:
+                        row[field] = table_row[field]
+                rows.append(row)
+
+        # Prepare query description
+        for field 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:
+            return rows
+
+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__()