Let's take a look, I'll just give you a couple examples and then I'll move to the actual [UNKNOWN] just so you can see things. So here's the mean. I took the mean of x, x doesn't exist. So if you take the mean of something that doesn't exist you're clearly going to get an error. Here it says so you get the error message this is produced by the stop function. And I call the traceback. It tells me where the error occurs. Well the error just occurs like right away at the top of the mean function. So that, so the function mean didn't actually call any other functions, right because it couldn't because it couldn't figure out, it couldn't evaluate your argument. And so, that's where the error occurs so, you know that may or may not be. Many times that's just that is where the error will occur at the top level function. so, that in that case the trace-back's not very handy. One thing that's useful to, that, that the trace-back is useful for, is for when you're communicating with someone else over email, for example, me. And you say, you know I get an error when I call, you know, whatever function. Its very useful if we just say use the trace-back, when I get after I get the error. And then I can see okay, wha, what was the list of function calls, I mean. Unless the trace-back looks like this, then near dud bug, but when you're communicating with someone about an error and a function, its very handy just to print out the trace-back, so that they can seek out where the error occurred in the hierarchy of functions. You have to call traceback immediately after the error occurs. If you execute some other code and then call traceback it's not going to work, right? Because the traceback will only give you the most recent error. If you execute another function and there's no error well then there's no traceback to give because there's no error in the, in the next. ex, execution. So you have to you've to call it right away. Here's a slightly more interesting trace back so I call it LM, which is the linear modeling function. I said y tilda x. So y tilda x. They don't exist. And so I'm getting an error that says y cannot be found. Were to that error occur? Well actually it occurred several levels deep actually. So here I called L M and the L M arrow function is a little funky. So I don't expect you to understand all this but L M called e val on what's called a mono frame. Then that, it's then and then that called e val again which called the model frame function which called the model frame default. Model frame is a generic function that called a model frame default function. Then it called the eval function again on this. And then, the and at the seventh level, here. That's where the error occurred. Okay. So and it basically it occurred because it tried to evaluate my formula here. Which is y tilda x. That's what all e val calls are for. And and it couldn't. And when it evaluated the formula it couldn't find the actual value of y and x. So then, that's where the error occurred. So printing out something like this can be very useful. If you're trying to get help from someone else and you're, and you're, and you're kind of trying to debug your function together, okay? So so that's the traceback, there. The debug function doesn't really work well in the static format like this, but I'll show, I'll just show you that, but But, basically I could debug the l m, so you can debug any function, doesn't matter if you wrote it or not. So I can debug the l m function, and now when I call l m y fill the x it you get this debugging in and I'll give you the the expression that you called. And I'll first thing it will do is it will print out the entire code of the function, so this is a lot longer than so I just I just put I just cut cut it off in the middle here but the first thing it does is it prints out the whole function body code okay. And then you get this little prompt at the bottom here the browse [UNKNOWN] so now you're in what's called the browser. Okay? And the browser is just like your r workspace, actually. you, you can think of it like a work space embedded within a work space. Right? So, and but, but the elements of the workspace. Are, sorry, so the environment of that workspace in the browser is your function environment. So the things that are in your function are what are in your environment. So at, at the very top of this function you just actually, there's nothing in your environment except for the function arguments, right. So what are the function arguments? Well, the first one is the formula? So I have a formula in my environment which is this y tilda x, but that's it. So anyway so now I can well there are actually I should say there are other arguments so l'm too which are default values. But they are not, they are not listed there, but they do, they also, they, I shouldn't say there is more than just a formula because there is other arguments. So here what I do is I press n for next and it runs the first line. When I hit n and then enter and it runs the first line. N then enter runs the second line. N enter third. I just keep hitting n, n, n, so I run, and I just execute each line until I get to the line where the error occurs. So when you get to the line where the error occurs. And you execute it. You'll get an error just like you did before. It's not going to magically fix the problem by itself. But, at least you'll know where that error occurred, okay? So and further more, suppose you want you can debug functions within the debugger, right? So, you can I can call, if I want so, I can call. So this match.call function I just kind of stepped right over it but I could have gone into this function by calling debug on match.call before executing it and then when this got executed I'd be in the third level of the browser. Ah, so these things can nest and then as you go so they come like a stack of browser frames right? So so you can, you can, you can call the debug function on functions. While you're, even while you're in the debugger. And you can step through functions even while you're kind of debugging your top level functions. The recover function, so the you can set the recover to be this kind of error handler by using the options function. So you say options error equals recover. And, and that will set a global option, right? So now this will, this will, this behavior will curve ever, everything you do as long as your r session is open. As soon as you quit it will go away and then if you start up R somewhere else it's not going, it's not going to save that option. So now I'm going to what I'm trying to do is I'm going to read that csv some file that doesn't exist. Okay, and then you get an error and you get the error message, cannot open the connection. And you get some warning also. And now bu, but instead of getting my console back, I get a little menu here. And this menu is the function call stack, okay. So it's the same thing that you would get back, if you called trace back after the error. Right, so reed.csv is that I called, remember reed.csv is just a [UNKNOWN] for reed.tables. So Reed.csv called reed.table and then reed.table called the file function, because we wanted to create a file connection with my file, no such file. Right? And then when it tried to create that file connection. Obviously you couldn't find the file, and that's where the error occurred, okay? So the error occurred at the third level in the function call stacks, so what you can do, is you can press a number one, two, three [SOUND] and then you can kind of browse the environment of that function. So if I want to see okay, what was the read, what was going on in read.csv, I can just press one here, and then it would show me kind of the environment of the read.csv. And, and then I can break out of that and look at the environment for read.table, and et cetera. And so you can, if you have a very long function call stack, you can kind of jump back and forth to see what was going on at each, at each function call. To to try out a pinpoint kind of where the problem occurs. Okay. Okay. So so just to summarize really quickly there's, there are basically three main indications of some sort of problem or condition. There's the message warning error and of the three only the error will stop execution of the function. When you're analyzing a function that has and you think has a problem make sure you can reproduce that problem. And then you make sure you can articulate how what do your expectations and how'd and what the output are that and how the output differs from what you were expecting or [LAUGH] I'd like to say what you are expecting. And so that the, the interactive tools trace back, debug, browser, trace, and recover can be used. And it's, and the keyword actually here is interactive. The, the reason these tools are useful is because they are interactive. You can kind of do things on the console. And of course the debugging tools are not a substitute for thinking. And so you should always think about writing your code before, you know, just throwing it to the wind and hoping the debugger will catch it.
