1
00:00:06,580 --> 00:00:09,060
- Correctness versus performance.

2
00:00:09,060 --> 00:00:11,940
So this is a big one for me,
because we've also been trained

3
00:00:11,940 --> 00:00:15,040
as developers to kinda
focus on performance, again,

4
00:00:15,040 --> 00:00:17,310
as a priority, and we tend to write code

5
00:00:17,310 --> 00:00:19,910
with the idea that this
gonna be faster than that.

6
00:00:19,910 --> 00:00:23,030
The reality is is that unless
you've done a thousand times

7
00:00:23,030 --> 00:00:25,470
and you already know, you're guessing,

8
00:00:25,470 --> 00:00:28,230
and usually code that's
focused on performance

9
00:00:28,230 --> 00:00:29,980
is clever code.

10
00:00:29,980 --> 00:00:31,750
I wanna reverse the tables on this.

11
00:00:31,750 --> 00:00:33,890
I don't wanna optimize for performance.

12
00:00:33,890 --> 00:00:37,090
I wanna optimize for correctness
when we're writing code,

13
00:00:37,090 --> 00:00:39,290
I wanna use our debuggers
and our profiles,

14
00:00:39,290 --> 00:00:41,940
not really our debuggers,
I apologize, our profiles,

15
00:00:41,940 --> 00:00:44,950
our ability to benchmark,
our ability to trace,

16
00:00:44,950 --> 00:00:48,060
to identify where we may
have a performance problem,

17
00:00:48,060 --> 00:00:50,700
and then fix the performance
issues like that,

18
00:00:50,700 --> 00:00:55,220
still trying to maintain levels
of correct or readable code.

19
00:00:55,220 --> 00:00:57,500
Look, the performance of your software

20
00:00:57,500 --> 00:00:59,440
is gonna come from four places,

21
00:00:59,440 --> 00:01:01,550
or I should say the lack of performance

22
00:01:01,550 --> 00:01:03,220
is gonna come from four places.

23
00:01:03,220 --> 00:01:06,680
First is latency, and it's
gonna be latency on networking

24
00:01:06,680 --> 00:01:09,630
and IO, disk IO, those types of things.

25
00:01:09,630 --> 00:01:12,190
That's going to be a real
killer of performance,

26
00:01:12,190 --> 00:01:14,850
and those are things you
probably want to focus on first.

27
00:01:14,850 --> 00:01:16,970
Your next potential performance strain

28
00:01:16,970 --> 00:01:18,690
is going to be on memory allocations.

29
00:01:18,690 --> 00:01:20,720
We're gonna talk about that in this class,

30
00:01:20,720 --> 00:01:23,970
and memory allocations can
cause performance problems,

31
00:01:23,970 --> 00:01:25,390
and we'll see that when we start

32
00:01:25,390 --> 00:01:26,730
talking about the garbage collector,

33
00:01:26,730 --> 00:01:28,930
and the garbage collector's
doing an amazing job,

34
00:01:28,930 --> 00:01:30,920
but we don't wanna walk away from it.

35
00:01:30,920 --> 00:01:32,540
When we talk about performance in GO,

36
00:01:32,540 --> 00:01:34,550
it's not about is it the fastest,

37
00:01:34,550 --> 00:01:36,600
it's about is it fast enough,

38
00:01:36,600 --> 00:01:39,470
but garbage collection
and memory allocations,

39
00:01:39,470 --> 00:01:41,160
that will be your number two.

40
00:01:41,160 --> 00:01:44,580
Number three is going to
be how you access data,

41
00:01:44,580 --> 00:01:46,630
how efficiently you access data.

42
00:01:46,630 --> 00:01:48,910
We're gonna talk about
that in this class as well.

43
00:01:48,910 --> 00:01:52,300
And then, number four would
be algorithm efficiencies,

44
00:01:52,300 --> 00:01:57,200
can we streamline an algorithm
to take less steps than more,

45
00:01:57,200 --> 00:02:00,120
and we tend to, if we're
optimizing performance,

46
00:02:00,120 --> 00:02:02,590
focus on that piece, and the reality is

47
00:02:02,590 --> 00:02:04,260
the hardware's gonna be so fast

48
00:02:04,260 --> 00:02:06,480
if we solve the first three problems,

49
00:02:06,480 --> 00:02:09,080
that we can have a little
algorithm inefficiency,

50
00:02:09,080 --> 00:02:12,980
and still get more than we
need in terms of performance,

51
00:02:12,980 --> 00:02:14,880
and now we're optimizing for correctness,

52
00:02:14,880 --> 00:02:17,460
and this is really where we wanna go.

53
00:02:17,460 --> 00:02:19,300
And there's some great quotes

54
00:02:19,300 --> 00:02:22,120
by some amazing people in this area.

55
00:02:22,120 --> 00:02:24,120
Wes Dyer, "Make it correct, make it clear,

56
00:02:24,120 --> 00:02:26,080
"make it concise, make it fast."

57
00:02:26,080 --> 00:02:28,450
I want us to listen to this history.

58
00:02:28,450 --> 00:02:30,250
These are people who've
already learned these lessons.

59
00:02:30,250 --> 00:02:33,430
I don't wanna make the same mistakes.

60
00:02:33,430 --> 00:02:36,180
Alright, JBD who works
over at Google, and again,

61
00:02:36,180 --> 00:02:37,250
you gotta look her up,

62
00:02:37,250 --> 00:02:41,000
she's also a tremendous
thought leader in this space.

63
00:02:41,000 --> 00:02:44,170
Engineering's not about
finding the perfect solutions,

64
00:02:44,170 --> 00:02:45,430
it's about understanding the trade-offs,

65
00:02:45,430 --> 00:02:46,610
we talked about this.

66
00:02:46,610 --> 00:02:49,090
Engineering's about understanding
the costs you're taking

67
00:02:49,090 --> 00:02:50,900
for the benefits that you need.

68
00:02:50,900 --> 00:02:53,750
It also is about
optimizing for correctness,

69
00:02:53,750 --> 00:02:56,230
knowing what we're doing
in terms of our code.

70
00:02:56,230 --> 00:02:58,750
And correctness of the implementation

71
00:02:58,750 --> 00:03:00,210
is the most important concern.

72
00:03:00,210 --> 00:03:01,930
Listen what Al is saying here,

73
00:03:01,930 --> 00:03:04,660
"but there's no royal
road", listen to Al again,

74
00:03:04,660 --> 00:03:07,000
what he's gonna say is that
it takes several things,

75
00:03:07,000 --> 00:03:09,590
invariants, that's the engineering costs,

76
00:03:09,590 --> 00:03:11,450
testing and code reviews.

77
00:03:11,450 --> 00:03:14,700
And testing and code reviews
really require our ability

78
00:03:14,700 --> 00:03:17,180
to read and understand code.

79
00:03:17,180 --> 00:03:19,700
Without our ability to
read and understand code,

80
00:03:19,700 --> 00:03:21,660
we're really going to be lost.

81
00:03:21,660 --> 00:03:24,190
There's a great quote that says that

82
00:03:24,190 --> 00:03:28,080
we are one of the few
industries where we ask people

83
00:03:28,080 --> 00:03:30,920
how to write code before
we teach them how to read,

84
00:03:30,920 --> 00:03:33,250
and this is really mind boggling

85
00:03:33,250 --> 00:03:34,940
when you start to think about it,

86
00:03:34,940 --> 00:03:37,190
because everything that
we're gonna talk about,

87
00:03:37,190 --> 00:03:38,560
about correctness,

88
00:03:38,560 --> 00:03:42,540
is going to stem from our
ability to read code first.

89
00:03:42,540 --> 00:03:47,030
And this class is all about
our ability to read code,

90
00:03:47,030 --> 00:03:50,910
to optimize for correctness,
to be productive.

91
00:03:50,910 --> 00:03:51,780
That's what we're trying to do,

92
00:03:51,780 --> 00:03:54,330
that's what GO is trying to do,

93
00:03:54,330 --> 00:03:57,240
and if we continue to look at
quotes even from Bryan, right,

94
00:03:57,240 --> 00:04:01,420
the basic ideas here of
good style, of correctness,

95
00:04:01,420 --> 00:04:04,640
they're fundamental to
writing clear and simple code,

96
00:04:04,640 --> 00:04:08,440
and we've known this, he's
known this for over 35 years.

97
00:04:08,440 --> 00:04:09,890
I'm not telling you anything new,

98
00:04:09,890 --> 00:04:12,230
these are not my ideas, these
are not my philosophies,

99
00:04:12,230 --> 00:04:15,300
I'm trying to bring to
you industry leaders

100
00:04:15,300 --> 00:04:17,170
who are telling us over and over again

101
00:04:17,170 --> 00:04:19,840
correctness over performance
when it comes to the code,

102
00:04:19,840 --> 00:04:24,210
and to do that, we've gotta
be very good at reading code.

103
00:04:24,210 --> 00:04:27,180
But I love what Jason Fried
said here, maybe again,

104
00:04:27,180 --> 00:04:29,350
next to Sarah May, one of
my favorite quotes here

105
00:04:29,350 --> 00:04:30,710
in the material,

106
00:04:30,710 --> 00:04:34,460
"Problems can usually be
solved with mundane solutions.

107
00:04:34,460 --> 00:04:36,280
"That means there's no glamorous work.

108
00:04:36,280 --> 00:04:38,120
"You don't get to show off."

109
00:04:38,120 --> 00:04:42,340
Okay, you build something,
it works, you move on, okay?

110
00:04:42,340 --> 00:04:44,270
And this is really important.

111
00:04:44,270 --> 00:04:45,103
This is one of those quotes

112
00:04:45,103 --> 00:04:47,420
I want you to put in
front of you at all times,

113
00:04:47,420 --> 00:04:51,120
because if you are a back-end
developer, like I am,

114
00:04:51,120 --> 00:04:53,950
and what you're looking to
become if you're not already,

115
00:04:53,950 --> 00:04:56,450
then the idea is that your software

116
00:04:56,450 --> 00:04:59,020
should be working 24/7 without issues.

117
00:04:59,020 --> 00:05:01,550
The reality is nobody
should know your name

118
00:05:01,550 --> 00:05:04,100
if you're doing your job well.

119
00:05:04,100 --> 00:05:05,710
If somebody knows your name,

120
00:05:05,710 --> 00:05:08,730
probably means your software's
failing or causing problems.

121
00:05:08,730 --> 00:05:10,020
You want people to know your name?

122
00:05:10,020 --> 00:05:11,470
Become a front-end dev.

123
00:05:11,470 --> 00:05:13,270
Front-end devs get the oohs and the aahs,

124
00:05:13,270 --> 00:05:14,540
but I promise you this,

125
00:05:14,540 --> 00:05:16,900
you're gonna spend two
weeks working on a screen,

126
00:05:16,900 --> 00:05:19,560
killing yourself to get
things in a certain location,

127
00:05:19,560 --> 00:05:21,700
certain words, and the first
thing somebody's gonna do

128
00:05:21,700 --> 00:05:23,780
when they look at it
is point something out

129
00:05:23,780 --> 00:05:25,030
that they don't like.

130
00:05:25,030 --> 00:05:26,770
Aah, that used to drive me crazy.

131
00:05:26,770 --> 00:05:28,480
I'm a back-end dev, okay?

132
00:05:28,480 --> 00:05:30,740
I build systems to the point

133
00:05:30,740 --> 00:05:32,750
where nobody knows I even exist.

134
00:05:32,750 --> 00:05:34,360
That's a good day in the office.