Add support for plugins

[niemyjski]

We should try and modularize the code base so the jquery/ajax and window.onerror logic is separated out. So if you're not using it, you're not paying the overhead cost.

[blackrabbit99]

I forked you library to start the experiment
I'm gonna try do it for my project but still have several questions about how I should maintain it

[niemyjski]

What are your questions :)? I'll work with you until it's perfect and then we'll help maintain it too via a pull request :D

[blackrabbit99]

So lets look on simple example. jQuery error handling.

jQuery.error = function (message) {
    // Inject TraceKit
};

The question number 1. We should discuss error format. Maybe is there reason to describe error format in separate file? Because It should be the same for all plugins and handling options.

The second question. Where should the plugins store? In separate repos or into some folder?

[niemyjski]

@blackrabbit99 you could take a look at how raven did it. Also, there is an existing pull request here: https://github.com/occ/TraceKit/pull/49/files

I also have been talking to someone on Skype who wanted to work on this too. I'll forward this to them.

[niemyjski]

@blackrabbit99, I kinda like the idea of a plugins/integrations folder, what about you? We'll currently I think our window.onerror script would be a good example.

For example newer browsers return an actual error object but if not they specify like 5 parameters and we resolve it from there into a stack trace..

I'd like to have one generic method that takes the original error if it's there and the stack trace maybe? I'd want the original error always because we could parse extra properties off it (something we currently can't do). But in the event of where we don't have the error object, we'd have to parse the stack trace from the given meta data and only pass the stack. Thoughts on this?

[fizerkhan]

@niemyjski That will be great. i started with Jquery plugin https://github.com/MindscapeHQ/raygun4js/blob/master/src/raygun.tracekit.jquery.js But i ended up with lot of anonymous and file 0 and col 0 in stacktraces during ajax errors. Do you have any idea why it is happening?

[niemyjski]

I'm not sure, but if you could create a small unit test / sample that reproduces I'll help take a look. I'm going to spend a good portion of today trying and finishing up the unit tests

[fizerkhan]

@niemyjski Created jsfiddle for this test case. I added the external resource for Jquery, Tracekit and Tracekit Jquery pluign. Please check it

http://jsfiddle.net/fizerkhan/w64j03r9/10/

[niemyjski]

@fizerkhan if you look at the thrown error there is no line numbers or anything other than the function name. With this said, do you think it could be because jquery is loaded from a cdn and there is a security / cors functionality that is blocking the script/browser?

[niemyjski]

@fizerkhan So I did some looking and pulled jquery from rawgit as well (http://jsfiddle.net/bniemyjski/hrh0y8kd/3/). I think it's because we are trying to guess the stack trace.. If you look at the passed in error, it's just a string.

What are your thoughts on this?

[fizerkhan]

@niemyjski Thanks. I too tried with crossorigin="anonymous" in Jquery script tag as well as with local Jquery. I get the same null valued stacktraces.

I don't have any idea why Jquery returns error message as string.

I have a question:

Why Tracekit returns stack traces of null file name, line number, and column number for an error message which is just a string.

[niemyjski]

@fizerkhan I think it's because it's trying to compute it from the current stack trace, there maybe some issues with this code. We'll need to take a look into this and debug your sample more.

[fizerkhan]

@niemyjski Thanks, we will do it. I really appreciate your quicker response. 👍