Message193325
The ElementTree module allows to write a XML parser using Python callbacks. The module relies on the expat library which is implemented in C. Expat calls these Python callbacks, but ElementTree does not check if a Python exception was raised or not.
Example 1:
-------------------
import unittest
from xml.etree import ElementTree as ET
class Target(object):
def start(self, tag, attrib):
print("start")
raise ValueError("raise start")
def end(self, tag):
print("end")
raise ValueError("raise end")
def close(self):
print("close")
raise ValueError("raise close")
parser = ET.XMLParser(target=Target())
parser.feed("<root><test /></root>")
-------------------
Output with Python 3.3:
-------------------
start
startendendTraceback (most recent call last):
File "x.py", line 18, in <module>
parser.feed("<root><test /></root>")
File "x.py", line 10, in end
print("end")
File "x.py", line 10, in end
print("end")
File "x.py", line 6, in start
print("start")
File "x.py", line 7, in start
raise ValueError("raise start")
ValueError: raise start
-------------------
start() was called twice, as end() method, even if the first start() method raised an exception.
The traceback is strange: it looks like end() was called by start(), which is wrong.
Example 2:
-------------------
import unittest
from xml.etree import ElementTree as ET
class Target(object):
def start(self, tag, attrib):
raise ValueError("raise start")
def end(self, tag):
raise ValueError("raise end")
def close(self):
raise ValueError("raise close")
parser = ET.XMLParser(target=Target())
parser.feed("<root><test /></root>")
-------------------
Output with Python 3.3:
-------------------
Traceback (most recent call last):
File "x.py", line 15, in <module>
parser.feed("<root><test /></root>")
File "x.py", line 9, in end
raise ValueError("raise end")
ValueError: raise end
-------------------
end() was called even if start() already failed. The exception which was set by start has been replaced by end() exception.
In my opinion, it's not a good thing to call PyEval_EvalFrameEx() and similar functions when a Python exception is set, because it behaves badly (ex: print("end") in Example 1 raises an exception... which is wrong, the traceback is also corrupted) and may replaces the old exception with a new exception (ex: "end" replaces "started"). |
|
Date |
User |
Action |
Args |
2013-07-18 21:01:01 | vstinner | set | recipients:
+ vstinner |
2013-07-18 21:01:01 | vstinner | set | messageid: <1374181261.54.0.8148762893.issue18501@psf.upfronthosting.co.za> |
2013-07-18 21:01:01 | vstinner | link | issue18501 messages |
2013-07-18 21:01:01 | vstinner | create | |
|