Interesting “getElementById() takes exactly 1 argument (2 given)”, sometimes it occurs. Can someone explain it?

Question

#-*- coding:utf-8 -*-
import win32com.client, pythoncom
import time

ie = win32com.client.DispatchEx(\'InternetExplorer.Application.1\')
ie.Visible = 1
ie.Navigate(\         


        

Answer 1

As a method of a COMObject, getElementById is built by win32com dynamically.
On my computer, if url is http://ieeexplore.ieee.org/xpl/periodicals.jsp, it will be almost equivalent to

def getElementById(self):
    return self._ApplyTypes_(3000795, 1, (12, 0), (), 'getElementById', None,)

If the url is www.baidu.com, it will be almost equivalent to

def getElementById(self, v=pythoncom.Missing):
    ret = self._oleobj_.InvokeTypes(1088, LCID, 1, (9, 0), ((8, 1),),v
            )
    if ret is not None:
        ret = Dispatch(ret, 'getElementById', {3050F1FF-98B5-11CF-BB82-00AA00BDCE0B})
    return ret

Obviously, if you pass an argument to the first code, you'll receive a TypeError. But if you try to use it directly, namely, invoke ie.Document.getElementById(), you won't receive a TypeError, but a com_error.

Why win32com built the wrong code?
Let us look at ie and ie.Document. They are both COMObjects, more precisely, win32com.client.CDispatch instances. CDispatch is just a wrapper class. The core is attribute _oleobj_, whose type is PyIDispatch.

>>> ie, ie.Document
(, >)
>>> ie.__class__, ie.Document.__class__
(,
 )
>>> oleobj = ie.Document._oleobj_
>>> oleobj

To build getElementById, win32com needs to get the type information for getElementById method from _oleobj_. Roughly, win32com uses the following procedure

typeinfo = oleobj.GetTypeInfo()
typecomp = typeinfo.GetTypeComp()
x, funcdesc = typecomp.Bind('getElementById', pythoncom.INVOKE_FUNC)
......

funcdesc contains almost all import information, e.g. the number and types of the parameters.
If url is http://ieeexplore.ieee.org/xpl/periodicals.jsp, funcdesc.args is (), while the correc funcdesc.args should be ((8, 1, None),).

Long story in short, win32com had retrieved the wrong type information, thus it built the wrong method.
I am not sure who is to blame, PyWin32 or IE. But base on my observation, I found nothing wrong in PyWin32's code. On the other hand, the following script runs perfectly in Windows Script Host.

var ie = new ActiveXObject("InternetExplorer.Application");
ie.Visible = 1;
ie.Navigate("http://ieeexplore.ieee.org/xpl/periodicals.jsp");
WScript.sleep(5000);
ie.Document.getElementById("browse_keyword").value = "Computer";

Duncan has already pointed out IE's compatibility mode can prevent the problem. Unfortunately, it seems it's impossible to enable compatibility mode from a script.
But I found a trick, which can help us bypass the problem.

First, you need to visit a good site, which gives us a HTML page, and retrieve a correct Document object from it.

ie = win32com.client.DispatchEx('InternetExplorer.Application')
ie.Visible = 1
ie.Navigate('http://www.haskell.org/arrows')
time.sleep(5)
document = ie.Document

Then jump to the page which doesn't work

ie.Navigate('http://ieeexplore.ieee.org/xpl/periodicals.jsp')
time.sleep(5)

Now you can access the DOM of the second page via the old Document object.

document.getElementById('browse_keyword').value = "Computer"

If you use the new Document object, you will get a TypeError again.

>>> ie.Document.getElementById('browse_keyword')
Traceback (most recent call last):
  File "", line 1, in 
TypeError: getElementById() takes exactly 1 argument (2 given)