Image via Wikipedia

Recently I came across an interesting problem (to say the least) where jsonrpc call would execute the error call back function, but did not provide the details.  Additionally (as the title suggests) Firebug showed that the call never completed. Firebug continues to show rotating ring next to this call. But server logs suggested that server not only received the call, but it also processed the request completely.

I googled for “jsonrpc never completes” (and other similar terms) But no one had faced the problem. Strange

Normally, when jsonrpc call executes error callback, there would be some error on the server (serving the rpc call) But in this case, the call went thru fine. No error on server.

So I started debugging. First place to look, check server logs. No error there. Added additional debug statements on server. Still no help.

Then I modified the client-side code. Checked if there were additional information in the response sent back. Nothing.

May be there was something wrong with this specific RPC. Just to isolate, I made another RPC, which was known to work at other places. I was so sure that problem would go away, that I was shocked that problem remained.

So clearly it wasn’t the RPC, but the way it was called. So I started looking at surrounding code. Wasn’t any different. So I went further up the stack. It turns out later in the chain – unconditional window.location.reload() was called. This was causing the browser to reload the page even before the jsonrpc would complete.

Looking back – it was obvious. So why did I write this post ? So that someone else who may come across such symptom in future, may have some tips to troubleshoot.

In case you are interested, this was part of this project

Related articles

• Debugging Javascript applications (markos.gaivo.net)
• Firebug Guide for Web Designers (sixrevisions.com)

I was trying to import the data from CSV file into sqlite3 database. Ideally this should be very simple task, with following the steps given in the sqlite tutorial. It is a matter of calling the sqlite command with separator argument, followed by an import operation, as listed below.

sqlite3 test.db  "create table t1 (t1key INTEGER PRIMARY KEY,data TEXT);"
sqlite3 -separator , test.db ".import some.csv t1"

Except that main attribute of my CSV file was that it could contain single records with embedded comma. I was hoping that sqlite3 would be smart enough to detect that the fields were enclosed within double quotes and then separate by comma. But I soon realized that only a code specifically dealing with CSV would know about this.  As we can see in the example above, the import is a generic code and as a user I listed comma as a separator.

My Data looked something like this :

"1","data1"
"2","data2,data3"

So like any *nix geek would do, I tried providing double quote and comma as a separator. To my surprise it worked very well. I though separator would take only single character, and I had provided two (three?). Anyway, important thing to remember is to escape the single quote with a backslash (I didn’t try it without the backslash, may be that would work too)

So here is the syntax that worked :

sqlite3 -separator \", test.db ".import mydata.csv mytbl"

Update : Turns out SQLite Manager is much better solution after all.  It is an Extension for Firefox and other apps to manage any sqlite database. Not only it took care of above situation, it also handled empty cells as well where the command line failed with following error message :

line 4: expected 3 columns of data but found 2

Data with missing cells : Notice two successive commas :

"1","data1","data2"
"2","data3,data4","data5"
"3",,"data6"

This is in continuation with my previous blog post :

To get IronPython to use Standard Python Modules,  one needs to add the following two lines to C:\IronPython-2.0.1\Lib\site.py :

import sys
sys.path.append(r"C:\Python25\Lib")

While this works for most part, it doesn’t help if you are using Python extensions written in C. More about my specific problems in another post. But there is an open source project IronClad to deal specifically with this issue. In the meantime, you can check the differences between IronPython and CPython

Reading (and writing to) CSV file is critical part of my program, while in stadard python it was as easy as “import csv”, the same thing took some efforts to get it working in IronPython. I got the following error for my import statement

Error on line 7 in csv.py
from functools import reduce

I also tried using ActiveState Python 2.5.2.2 (which I already had from few months ago, didn’t feel like downloading the latest version till the problem was fixed) But that didn’t help either. With ActiveState, I got the same error on the same line, except this time it was for _csv.

To Quote from IronPython Cookbook :

For some reason the Python standard library csv module is written in C, which means that it isn’t available to IronPython.

The cookbook points to a third party library called A Fast Csv Reader . The cookbook has a nice example of how to use the said DLL with your IronPython Program.

It wasn’t clear to me as to why I had to register at Code Project to download this binary since it is provided under MIT Open Source License. But whom am I gonna complain to ? Beggars can’t be choosers

I used the subprocess module of python to invoke the external program. It was easy to capture the output sent on stdout. In order to send ‘y‘ on stdin for each iteration,  I tried sending ‘y‘ on stdin, but that would not work.  The script would hang. After discussing this, with more experienced python programmer, it was suggested that one possible reason why the script hangs is may be because the stdin buffer wasn’t flushed.  Both of us were not sure how to do that. Then it was discussed that when we run the external program from command line, we not only type ‘y‘ as response, but we also hit Enter there after, which results in flushing the stdin buffer. So may be that is what I ought to try.

To my surprise, it worked. So the solution was to pass ‘y/n‘ instead of single ‘y‘

Here is how my code looked like :

import subprocess
cmd = "...." # The command you wish to execute
proc = subprocess.Popen(cmd,
                       stdin=subprocess.PIPE,
                       stdout=subprocess.PIPE)
print proc.communicate('Y\n')[0]


Here is the quick list of pdb commands

• n : Execute next line
• s : Step inside a module
• p : Print the value of a variable
• c : Continue
• l : List. Shows stack trace
• q : Quit (Most important command )

For details, check this site

But what if I can’t debug interactively ?

Typically, you come across this problem when you are running the script on a server. Even if the script worked OK in test environment, in live environment something goes wrong, because the script is executed by server, and you can not control the arguments that are being passed to the script. In this scenario, you need to resort to age-old logging technique

Here are few line of code you need to add to your script. Modify as per your needs (especially paths)
Earlier in your script (as early as possible) add following lines :

from time import strftime
f = open('D:/temp/log.txt', 'r+')
f.write("-- Begin Log at %s --n" % strftime("%Y-%m-%d %H:%M:%S"))


At the end of your script (as late as possible) add following lines :

f.write("-- End Log at %s --n" % strftime("%Y-%m-%d %H:%M:%S"))
f.close()


Your basic logging infrastructure is ready. Now you can add f.write() lines through out your code, as and where needed.

Note : I’m part of pythondev group on twitter.  Visit the website for details.