1 00:00:05,360 --> 00:00:08,880 We left the previous video with three different ways to create the same output, 2 00:00:08,950 --> 00:00:11,770 and the question: "Which way performs better?". 3 00:00:11,770 --> 00:00:14,520 So the code's in compchallenge2b.py 4 00:00:14,740 --> 00:00:20,200 But before we start comparing how fast the three approaches are, we should make sure we're comparing 5 00:00:20,440 --> 00:00:21,810 like for like. 6 00:00:21,880 --> 00:00:28,840 And in this case we're not. Now our "nested for loops" creates six lists, and discards each list before creating 7 00:00:28,840 --> 00:00:29,580 the next one. 8 00:00:29,650 --> 00:00:34,450 And you can see, here on line 18, that new list being created each time through the loop. 9 00:00:34,450 --> 00:00:39,670 Now that "List comprehension inside a for loop", this next section of code, that does the same thing each 10 00:00:39,670 --> 00:00:40,420 er each 11 00:00:40,450 --> 00:00:44,240 time around the loop, the comprehension assigns a new list to 12 00:00:44,240 --> 00:00:47,200 exits_to_destination_2 13 00:00:47,300 --> 00:00:48,970 You can say that happening on line 30. 14 00:00:49,300 --> 00:00:55,910 And thirdly, the "nested comprehension" builds up a nested list containing all 6 of the other lists. 15 00:00:55,980 --> 00:00:59,350 This is the code down here, starting at line 36. 16 00:00:59,410 --> 00:01:05,200 So before we can perform any meaningful analysis, we need to decide what we're trying to measure here. 17 00:01:05,440 --> 00:01:10,630 Now if we want to use the lists somewhere else in our code, then we're really interested in how efficiently 18 00:01:10,690 --> 00:01:16,680 the lists can be created - and only the third method is creating the complete list here. 19 00:01:17,080 --> 00:01:22,170 But on the other hand, if we're only interested in displaying the result, then we can measure the code 20 00:01:22,170 --> 00:01:23,840 as it is at the moment. 21 00:01:23,830 --> 00:01:26,040 Now as we're learning about all this we're going to do both 22 00:01:26,040 --> 00:01:31,810 here. We'll start by measuring how quickly the results can be displayed, because that's what the code's doing 23 00:01:31,810 --> 00:01:32,530 at the moment. 24 00:01:32,710 --> 00:01:36,550 But there is one change we have to make, to keep the comparison fair. 25 00:01:36,550 --> 00:01:39,710 So our nested loop is printing the list twice. 26 00:01:39,790 --> 00:01:45,670 So I'm going to start here by looking at line 40, and actually delete those three lines. So printing out the 27 00:01:45,670 --> 00:01:47,460 exits to Destination 3. 28 00:01:47,740 --> 00:01:54,760 And also that print there as well. So I'm actually going to delete those lines. Like so. And that's just again to make 29 00:01:54,760 --> 00:01:57,690 sure that we're keeping the comparisons fair here. 30 00:01:57,910 --> 00:02:00,640 All right so let's have a look at the timeit module. 31 00:02:01,120 --> 00:02:02,030 I'll open up a browser. 32 00:02:07,250 --> 00:02:08,949 This is a good place to start. 33 00:02:08,960 --> 00:02:13,510 Now if you want to run tests from the command line, the documentation shows how to use the timeit 34 00:02:13,510 --> 00:02:14,880 module that way. 35 00:02:14,990 --> 00:02:20,330 So that might be fine for testing a single line Python statement, but it gets increasingly complex when 36 00:02:20,330 --> 00:02:23,320 testing larger or longer blocks of code. 37 00:02:23,330 --> 00:02:28,730 Remember the indentation's vitally important in Python, which means you need to start multiple strings 38 00:02:28,820 --> 00:02:31,390 with the correct number of spaces. 39 00:02:31,400 --> 00:02:36,110 So by all means experiment with a command line interface if you want, but we won't be doing things that 40 00:02:36,110 --> 00:02:36,700 way. 41 00:02:37,040 --> 00:02:43,370 So to use timeit in our code, we have to import the timeit module, then either use the timeit function or 42 00:02:43,370 --> 00:02:45,870 create an instance of the Timer class. 43 00:02:45,890 --> 00:02:51,080 Now the timeit function is one of three convenience functions that make use of the Timer class a bit 44 00:02:51,080 --> 00:02:51,750 easier. 45 00:02:52,010 --> 00:02:54,490 Although there's just not much difference in the code you have to write. 46 00:02:54,590 --> 00:02:59,290 So I'm going to use the timeit function, because printing the result is slightly easier. 47 00:02:59,660 --> 00:03:08,290 So coming down here to 27.5.2, the timeit function is passed a statement to 48 00:03:08,290 --> 00:03:12,870 execute. And we're going to start off by passing a statement as a string, 49 00:03:13,000 --> 00:03:15,320 but you can also pass a function instead. 50 00:03:15,370 --> 00:03:18,220 As long as the function doesn't take any arguments. 51 00:03:18,600 --> 00:03:19,750 So that's the first one. 52 00:03:19,770 --> 00:03:25,570 The first statement argument there. The next one is a uh the next parameter is a string containing 53 00:03:25,570 --> 00:03:26,310 setup code. 54 00:03:26,360 --> 00:03:28,200 We're going to come back to that in a moment. 55 00:03:28,210 --> 00:03:34,620 Now the default timer, you can see over here, the third argument, that can be changed if you feel the need to. 56 00:03:34,690 --> 00:03:37,670 Although there is very rarely any need to use a different timer. 57 00:03:38,140 --> 00:03:44,320 Now we looked at the time.perf_counter function in an earlier section, when we were investigating 58 00:03:44,320 --> 00:03:49,180 the time module. That's now the default timer, since Python 3.3 and as I said, there's 59 00:03:49,180 --> 00:03:51,090 rarely any need to change that. 60 00:03:51,180 --> 00:03:56,110 Now this fourth argument, number, is how many times to execute the code. 61 00:03:56,170 --> 00:04:00,970 Now the default value is, as you can see there, is 1 million; which means your code will be executed one million 62 00:04:00,970 --> 00:04:02,120 times. 63 00:04:02,140 --> 00:04:05,530 Now that might seem like a lot, but it does give good results. 64 00:04:05,860 --> 00:04:09,510 But one thing we probably shouldn't do though, is leave it set to the default 65 00:04:09,520 --> 00:04:14,500 when our code prints things out. Now at the moment, each of our methods prints the lists, and printing is 66 00:04:14,500 --> 00:04:16,120 quite a slow operation. 67 00:04:16,120 --> 00:04:20,700 Now if we leave the default number there, it will take about 1 minute to execute each block of code 1 68 00:04:20,709 --> 00:04:22,079 million times. 69 00:04:22,150 --> 00:04:28,260 And lastly this global argument, the final parameter, that's used to specify the namespace that our code 70 00:04:28,270 --> 00:04:32,820 will run in. Now I'm going to come back to that when we look at the setup string, because it's easier to explain 71 00:04:32,830 --> 00:04:35,950 by seeing an example. Alright, so that's the timeit function. 72 00:04:35,950 --> 00:04:37,690 Let's get back to our code. 73 00:04:38,430 --> 00:04:43,270 Let's start setting this up, so first things that we need to do is import the timeit module. 74 00:04:43,570 --> 00:04:44,950 So we'll do that, obviously, on line one 75 00:04:47,780 --> 00:04:51,300 Ok, so import timeit. 76 00:04:51,610 --> 00:04:57,030 Right, so at this point we've now got a choice; and which one you choose to do is entirely up to you. 77 00:04:57,050 --> 00:05:02,540 Now we can either turn our three sections of code into functions or we can wrap them in quotes to turn 78 00:05:02,540 --> 00:05:03,830 them into strings. 79 00:05:03,830 --> 00:05:08,400 Now I'm going to use strings, because turning each code block into a function's easy. 80 00:05:08,510 --> 00:05:11,960 Now there's a couple of things to watch out for when using strings. 81 00:05:11,990 --> 00:05:14,100 First, use triple quotes. 82 00:05:14,270 --> 00:05:18,260 That way any reformatting of your code can be kept to a minimum. 83 00:05:18,290 --> 00:05:19,760 So let's go down and start running some code. 84 00:05:19,760 --> 00:05:22,280 I'm going to call the first string nested_loop. 85 00:05:22,460 --> 00:05:35,440 So I'm goint to start by putting some code here, on line 19. So we've got nested_loop equals and then three quotes and backslash. 86 00:05:35,550 --> 00:05:43,210 Then I'm going to come down here, to line 27, and I put the three quotes again. Like so. And again, notice 87 00:05:43,250 --> 00:05:47,250 the backslash at the end of the opening quotes, here on line 19. 88 00:05:47,380 --> 00:05:52,420 We don't really want our code to start with a blank line, and the line continuation character avoids 89 00:05:52,420 --> 00:05:53,290 that. 90 00:05:53,290 --> 00:05:58,330 Now be careful when typing in code, rather than converting existing code to a string. 91 00:05:58,330 --> 00:05:59,710 The code must start with 92 00:05:59,720 --> 00:06:04,910 no indentation, then be correctly indented relative to the first line. 93 00:06:04,990 --> 00:06:11,440 Incorrect indentation causes an error when timeit tries to execute the code, but because it's a string 94 00:06:11,800 --> 00:06:13,750 you won't get any help from your IDE. 95 00:06:14,090 --> 00:06:14,370 OK. 96 00:06:14,410 --> 00:06:16,490 So that first bit, that's fairly straightforward. 97 00:06:16,510 --> 00:06:19,980 We've now got a string containing our first block of code. 98 00:06:20,260 --> 00:06:23,580 So we want to do the same thing with the other bits of code that we want to use in our test. 99 00:06:23,920 --> 00:06:29,320 So for the "list comprehension inside a for loop" then it'll go below the prints there. 100 00:06:29,730 --> 00:06:37,420 Well call it loop_comp equals, three double quotes again, and a backslash. 101 00:06:37,420 --> 00:06:41,570 Then we go down to just before the print and we put three double quotes. 102 00:06:42,180 --> 00:06:43,580 OK so that's our second one. 103 00:06:43,780 --> 00:06:44,980 And our third one, for the nested 104 00:06:44,990 --> 00:06:48,100 comprehension, below the two prints again. 105 00:06:48,430 --> 00:06:56,170 We're going to do nested_comp equals three double quotes again, backslash, just scroll down 106 00:06:56,260 --> 00:06:57,180 a little bit. 107 00:06:57,400 --> 00:07:01,400 Then we want to do three double quotes after that for loop. 108 00:07:02,410 --> 00:07:09,280 OK. So we've now, at this point, got three strings, that we can pass to the timeit function to see how their performance 109 00:07:09,280 --> 00:07:10,700 compares. 110 00:07:10,750 --> 00:07:11,840 Now we're going to get an error. 111 00:07:11,860 --> 00:07:15,060 But I'm going to go ahead anyway, because it's a good way to explain about 112 00:07:15,070 --> 00:07:19,710 those setup and globals parameters, that we saw in the documentation earlier. 113 00:07:19,770 --> 00:07:24,920 So what I'm going to do is, on line 50, we're going to put a bit of code in here. 114 00:07:24,950 --> 00:07:27,400 I'm going to put or type, result_1 115 00:07:27,400 --> 00:07:34,800 result_1 is equal to timeit.timeit and in parentheses we're going to put nested_loop. 116 00:07:38,230 --> 00:07:47,470 And then we're going to do print("Nested loop colon backslash T for a tab character, 117 00:07:47,500 --> 00:07:55,120 left and right curly braces, ending double quote, dot format and in parentheses result_1, and closing 118 00:07:55,120 --> 00:07:59,230 off the two parentheses. Blank line on line 52. 119 00:07:59,500 --> 00:08:00,670 So if we now run this 120 00:08:05,450 --> 00:08:06,560 Scroll up and have a look at the top 121 00:08:10,310 --> 00:08:17,210 and you can see we've got some errors, but right down at the bottom we've got this NameError: name 'locations' is not defined. 122 00:08:17,240 --> 00:08:23,130 Now that seems strange, because we can clearly see our locations dictionary, if we go up to the top here 123 00:08:23,290 --> 00:08:26,620 on line 3, that's been defined, and it is at the top of the file. 124 00:08:26,900 --> 00:08:32,720 But the problem is that timeit isn't aware of anything defined in your program, except for the arguments 125 00:08:32,720 --> 00:08:34,600 we pass to it, of course. 126 00:08:34,610 --> 00:08:40,370 So that's why it provides the setup and globals parameters, to allow us to set up the environment that 127 00:08:40,370 --> 00:08:41,960 our code will execute in. 128 00:08:41,960 --> 00:08:47,850 Now I'd normally use setup when possible, as it allows you to be more specific about what you pass 129 00:08:47,850 --> 00:08:48,370 to it. 130 00:08:48,630 --> 00:08:53,330 So I'm going to start with globals here though, because it's a simple change to the code. And once you've seen 131 00:08:53,330 --> 00:08:58,130 that working, I'll change the code to provide the dictionaries in the setup argument. 132 00:08:58,230 --> 00:09:00,980 We've already seen how to get the global variables for our modules, 133 00:09:01,070 --> 00:09:03,380 earlier in the course, using the globals() function. 134 00:09:03,380 --> 00:09:09,930 So we can pass the result of calling globals() as the globals argument to timeit. so I'm going to do that, and go down to 135 00:09:09,930 --> 00:09:12,090 line 50 again. I'll close the Run Window first 136 00:09:14,930 --> 00:09:19,610 And here, what we're going to do, on line 50, we'll actually pass that. So we're on the next line 137 00:09:19,610 --> 00:09:28,100 after nested_loop comma globals equals, then a call to the globals function: globals(). 138 00:09:28,360 --> 00:09:34,610 So our code snippet will now execute in our global namespace, which means that everything defined in our module 139 00:09:34,760 --> 00:09:36,860 will be available to the snippet. 140 00:09:37,040 --> 00:09:42,760 Now, that could well be overkill, for testing a small snippet of code, but it can be useful if the environment's 141 00:09:42,770 --> 00:09:45,710 too complex to set up in a small block of code. 142 00:09:46,100 --> 00:09:50,930 So there's another reason I prefer not to use globals to set the namespace, and that's because it was 143 00:09:50,930 --> 00:09:56,730 introduced in Python 3.5, and so therefore it won't work in earlier versions. 144 00:09:57,300 --> 00:10:01,200 So what we've done here will work, but I'm not going to run it right away. 145 00:10:01,640 --> 00:10:05,810 And that's because the default number of times for timeit is 1 million runs of the snippet, which we 146 00:10:05,810 --> 00:10:08,850 talked about earlier, and that would take about a minute to finish. 147 00:10:09,020 --> 00:10:10,900 And that's a long time to stare at the screen. 148 00:10:10,910 --> 00:10:14,550 I'm going to reduce the number to 1000 before running the program. 149 00:10:14,650 --> 00:10:23,980 So to do that, we're going to put a comma after the parentheses, and number=1000. 150 00:10:24,000 --> 00:10:27,140 Now just to be clear, I could've left that set to the default. 151 00:10:27,270 --> 00:10:29,160 And you'd normally want to do that. 152 00:10:29,250 --> 00:10:32,950 I just don't want to spend a minute in silence on the video, while we wait for it to finish. 153 00:10:33,290 --> 00:10:33,900 Let's run that 154 00:10:38,050 --> 00:10:43,500 And you can see, down the bottom, that it's printed out the time it took for our snippet to be executed 1000 times. 155 00:10:43,540 --> 00:10:46,230 Now the exact time will depend on your processor speed, 156 00:10:46,240 --> 00:10:51,720 how much memory your computer's got, and all sorts of other factors. Python version, for example. And if you've 157 00:10:51,730 --> 00:10:56,330 got other programs running, that will also have an impact on how quickly the code executes. 158 00:10:56,540 --> 00:11:01,150 It's likely that your computer's streaming this video, and doing that will certainly be taking up a fair 159 00:11:01,150 --> 00:11:03,340 bit of your processor's time as well. 160 00:11:03,540 --> 00:11:08,580 Now my Mac's busy recording this video, at the same time as I'm performing these timing tests. 161 00:11:08,620 --> 00:11:12,360 So these aren't really ideal conditions to be doing this, for either of us. 162 00:11:12,430 --> 00:11:16,960 I'd be very surprised, frankly, if you got the exact same value as me; but that's actually not what the 163 00:11:16,960 --> 00:11:18,390 timeit module's for. 164 00:11:18,580 --> 00:11:25,290 It's used to compare different code snippets, not to work out exactly how long your code took to execute. So 165 00:11:25,350 --> 00:11:30,280 running different snippets on the same computer will give a good indication of which one is the fastest. 166 00:11:30,640 --> 00:11:36,760 But be aware of what else your computer is doing at the same time, and ideally close down any other programs 167 00:11:36,760 --> 00:11:42,060 that aren't essential to the test. Alright, so let's finish the video here. In the next video, 168 00:11:42,060 --> 00:11:47,400 We'll start looking at that setup argument, which I talked about as an alternative, to fix that error we got 169 00:11:47,400 --> 00:11:48,460 with locations. 170 00:11:48,630 --> 00:11:50,250 So I'll see you in the next video.