1
00:00:00,480 --> 00:00:03,480
foreign

2
00:00:12,980 --> 00:00:16,379
giving a talk on the complexity of

3
00:00:15,120 --> 00:00:21,060
simplicity

4
00:00:16,379 --> 00:00:23,279
uh Chris used to run pycon IU and um

5
00:00:21,060 --> 00:00:24,600
has passed those rains over well as he

6
00:00:23,279 --> 00:00:26,820
has moved to the U.S he is the director

7
00:00:24,600 --> 00:00:28,320
of the Python software Foundation a

8
00:00:26,820 --> 00:00:31,880
contributor to pants which we're all

9
00:00:28,320 --> 00:00:31,880
wearing of course wait what

10
00:00:32,300 --> 00:00:36,899
and um yes probably the person one of

11
00:00:35,820 --> 00:00:38,219
the people at the conference needs the

12
00:00:36,899 --> 00:00:40,559
least amount of introduction because you

13
00:00:38,219 --> 00:00:43,379
all know him so well please help me give

14
00:00:40,559 --> 00:00:44,879
a warm welcome to Chris oh no an

15
00:00:43,379 --> 00:00:47,340
audience

16
00:00:44,879 --> 00:00:49,320
I was fearing this

17
00:00:47,340 --> 00:00:51,360
um hi everyone

18
00:00:49,320 --> 00:00:53,100
um this is another one of my talks about

19
00:00:51,360 --> 00:00:55,739
how pythonisters think about software

20
00:00:53,100 --> 00:00:57,780
engineering I've been giving talks in

21
00:00:55,739 --> 00:01:00,079
this vague direction for like five years

22
00:00:57,780 --> 00:01:00,079
now

23
00:01:01,260 --> 00:01:04,799
um and because I've been doing this for

24
00:01:03,000 --> 00:01:07,200
five years things are now starting to

25
00:01:04,799 --> 00:01:09,479
get scarily abstract which I apologize

26
00:01:07,200 --> 00:01:11,460
uh there is less code in this talk than

27
00:01:09,479 --> 00:01:13,500
my talk tomorrow will have

28
00:01:11,460 --> 00:01:15,180
um unfortunately uh for those of you

29
00:01:13,500 --> 00:01:17,460
seen other talks where I make old

30
00:01:15,180 --> 00:01:19,320
versions of python do weird things I'm

31
00:01:17,460 --> 00:01:20,580
not doing any of that today I I

32
00:01:19,320 --> 00:01:22,680
apologize

33
00:01:20,580 --> 00:01:24,180
um but you're here anyway um thank you

34
00:01:22,680 --> 00:01:26,280
so much for coming along

35
00:01:24,180 --> 00:01:30,180
Shall We Begin

36
00:01:26,280 --> 00:01:33,060
so here is a value of pythons simple is

37
00:01:30,180 --> 00:01:35,340
better than complex and what better way

38
00:01:33,060 --> 00:01:37,140
to demonstrate Python's Simplicity than

39
00:01:35,340 --> 00:01:38,460
with the first python code that most

40
00:01:37,140 --> 00:01:40,860
people write

41
00:01:38,460 --> 00:01:43,979
let's talk about print

42
00:01:40,860 --> 00:01:46,259
uh print it is the first function I see

43
00:01:43,979 --> 00:01:49,020
introduced in most introductory python

44
00:01:46,259 --> 00:01:52,020
textbooks understanding print is very

45
00:01:49,020 --> 00:01:54,479
very simple you call it unlike on this

46
00:01:52,020 --> 00:01:56,399
slide you normally give it an argument

47
00:01:54,479 --> 00:01:59,759
and then some text gets displayed to the

48
00:01:56,399 --> 00:02:02,820
screen uh wonderful but what does print

49
00:01:59,759 --> 00:02:06,259
do who here knows what print does

50
00:02:02,820 --> 00:02:08,340
print does a lot of error checking

51
00:02:06,259 --> 00:02:10,020
here's the actual implementation of

52
00:02:08,340 --> 00:02:12,780
print it's taken from the very source of

53
00:02:10,020 --> 00:02:15,120
C python the function is called built-in

54
00:02:12,780 --> 00:02:17,040
print impul which is complexity that you

55
00:02:15,120 --> 00:02:18,480
didn't need to know about that's the

56
00:02:17,040 --> 00:02:20,280
function that actually gets called it's

57
00:02:18,480 --> 00:02:23,459
quite long there is a lot of error

58
00:02:20,280 --> 00:02:24,959
checking here is the line somewhere in

59
00:02:23,459 --> 00:02:27,120
it that actually takes an argument

60
00:02:24,959 --> 00:02:28,400
writes it to what will usually be

61
00:02:27,120 --> 00:02:31,500
standard out

62
00:02:28,400 --> 00:02:33,780
and the bit that's really important is

63
00:02:31,500 --> 00:02:36,080
this bit here the bit that says Pi file

64
00:02:33,780 --> 00:02:38,760
underscore right object

65
00:02:36,080 --> 00:02:41,040
which itself has a whole world of

66
00:02:38,760 --> 00:02:44,220
implementation details to to worry about

67
00:02:41,040 --> 00:02:45,599
and you're looking at this and what's

68
00:02:44,220 --> 00:02:47,459
actually important about this is not

69
00:02:45,599 --> 00:02:50,599
none of none of what you're reading

70
00:02:47,459 --> 00:02:56,459
the important thing is that this is C

71
00:02:50,599 --> 00:02:58,800
and this is a python conference and so

72
00:02:56,459 --> 00:03:01,200
for you and me as python people

73
00:02:58,800 --> 00:03:02,879
understanding the C API is this going to

74
00:03:01,200 --> 00:03:04,560
add an undesirable amount of complexity

75
00:03:02,879 --> 00:03:05,280
to this talk

76
00:03:04,560 --> 00:03:06,599
um

77
00:03:05,280 --> 00:03:08,340
but I'd like to take a moment to point

78
00:03:06,599 --> 00:03:10,440
out here that this is already complexity

79
00:03:08,340 --> 00:03:12,540
that's inherent in doing something very

80
00:03:10,440 --> 00:03:14,159
very simple in Python rather than just

81
00:03:12,540 --> 00:03:17,519
executing some code that's written in

82
00:03:14,159 --> 00:03:20,340
Python we are executing C code python

83
00:03:17,519 --> 00:03:22,620
needs to allow you to run C functions

84
00:03:20,340 --> 00:03:24,239
and to keep that overall Simplicity it

85
00:03:22,620 --> 00:03:26,159
needs to prevent uh present these C

86
00:03:24,239 --> 00:03:27,900
functions to the rest of the language

87
00:03:26,159 --> 00:03:30,120
runtime in exactly the same way that

88
00:03:27,900 --> 00:03:31,980
presents your python functions and the

89
00:03:30,120 --> 00:03:33,659
CIA API that's needed to facilitate all

90
00:03:31,980 --> 00:03:35,760
that is complex you saw all that error

91
00:03:33,659 --> 00:03:37,500
handling but this is a python talk so

92
00:03:35,760 --> 00:03:40,379
let's talk about python

93
00:03:37,500 --> 00:03:42,980
print to a python developer has a

94
00:03:40,379 --> 00:03:45,180
complex interface with complex behaviors

95
00:03:42,980 --> 00:03:47,220
that really depend on how you specify

96
00:03:45,180 --> 00:03:48,599
these keyword arguments at the end and

97
00:03:47,220 --> 00:03:49,920
that's complexity that I don't want to

98
00:03:48,599 --> 00:03:52,440
touch so we're going to narrow the scope

99
00:03:49,920 --> 00:03:56,040
even more let's talk about printing just

100
00:03:52,440 --> 00:03:58,140
one thing just a single argument

101
00:03:56,040 --> 00:04:00,599
so what we're going to do is we're going

102
00:03:58,140 --> 00:04:01,680
to Define a variable and then we're

103
00:04:00,599 --> 00:04:04,140
going to print the value of that

104
00:04:01,680 --> 00:04:07,140
variable and if we set the value of our

105
00:04:04,140 --> 00:04:07,680
variable to a string like Hello World

106
00:04:07,140 --> 00:04:10,260
um

107
00:04:07,680 --> 00:04:12,180
this is very simple uh how would print

108
00:04:10,260 --> 00:04:13,980
work well we can we can iterate over

109
00:04:12,180 --> 00:04:16,380
each of the characters in The String we

110
00:04:13,980 --> 00:04:18,720
can write each character to a file stand

111
00:04:16,380 --> 00:04:20,100
it out in sequence

112
00:04:18,720 --> 00:04:21,540
um it doesn't actually work like this

113
00:04:20,100 --> 00:04:23,040
under the hood you saw that earlier

114
00:04:21,540 --> 00:04:25,199
there's some complexity that involves

115
00:04:23,040 --> 00:04:27,180
block i o which becomes more efficient

116
00:04:25,199 --> 00:04:29,759
if you transfer in larger blocks across

117
00:04:27,180 --> 00:04:30,900
the the API boundary but event there's

118
00:04:29,759 --> 00:04:31,919
eventually there's something that's just

119
00:04:30,900 --> 00:04:34,380
going to go through every single

120
00:04:31,919 --> 00:04:36,240
character and put them onto some sort of

121
00:04:34,380 --> 00:04:38,520
buffer somewhere so there's already a

122
00:04:36,240 --> 00:04:40,919
whole bunch of complex conceptual

123
00:04:38,520 --> 00:04:41,880
complexity here but it's it's not too

124
00:04:40,919 --> 00:04:44,759
bad

125
00:04:41,880 --> 00:04:46,680
as a python person what if we want to

126
00:04:44,759 --> 00:04:50,340
print an integer

127
00:04:46,680 --> 00:04:52,500
well this code is going to it's going to

128
00:04:50,340 --> 00:04:55,380
throw an index error because you can't

129
00:04:52,500 --> 00:04:57,300
iterate over an in object this code this

130
00:04:55,380 --> 00:04:58,680
code won't work

131
00:04:57,300 --> 00:05:00,360
um

132
00:04:58,680 --> 00:05:02,820
so what you need to do is you need you

133
00:05:00,360 --> 00:05:05,460
need to stringify the the parameter that

134
00:05:02,820 --> 00:05:07,620
we pass in to print

135
00:05:05,460 --> 00:05:09,380
and under the hood what stringification

136
00:05:07,620 --> 00:05:11,639
will look something like this you you

137
00:05:09,380 --> 00:05:13,139
dereference the type of the object and

138
00:05:11,639 --> 00:05:15,240
you call the dunder string method that's

139
00:05:13,139 --> 00:05:16,199
on that type pass in the parameter that

140
00:05:15,240 --> 00:05:17,940
was part

141
00:05:16,199 --> 00:05:19,320
and with that you know you can print

142
00:05:17,940 --> 00:05:22,560
anything you can print anything recently

143
00:05:19,320 --> 00:05:24,240
but in reality this this stringify code

144
00:05:22,560 --> 00:05:25,919
is probably provided us to the C library

145
00:05:24,240 --> 00:05:28,440
but if we're not going to use that and

146
00:05:25,919 --> 00:05:29,820
to deal with this complexity we're going

147
00:05:28,440 --> 00:05:31,199
to write our own string of fire which

148
00:05:29,820 --> 00:05:33,419
which needs a bunch of knowledge of

149
00:05:31,199 --> 00:05:35,160
ASCII and stuff and

150
00:05:33,419 --> 00:05:37,259
that's before you even get to that

151
00:05:35,160 --> 00:05:38,699
implementation of this put C function I

152
00:05:37,259 --> 00:05:39,840
didn't even talk about which is

153
00:05:38,699 --> 00:05:41,100
underpinned by a whole bunch of

154
00:05:39,840 --> 00:05:43,139
sophisticated software and Hardware

155
00:05:41,100 --> 00:05:45,120
Concepts that'll eventually give you the

156
00:05:43,139 --> 00:05:48,170
ability to see the text that will you

157
00:05:45,120 --> 00:05:51,629
asked to be displayed on your screen

158
00:05:48,170 --> 00:05:51,629
[Applause]

159
00:05:55,759 --> 00:06:01,199
nobody sets out to write complex code

160
00:05:59,639 --> 00:06:02,639
there's two things that you should have

161
00:06:01,199 --> 00:06:04,680
learned from that discussion of print

162
00:06:02,639 --> 00:06:06,300
here the first is that we should all

163
00:06:04,680 --> 00:06:08,400
want to interact with the simplest

164
00:06:06,300 --> 00:06:09,780
interface possible and the second is

165
00:06:08,400 --> 00:06:11,940
that even the simplest thing you can

166
00:06:09,780 --> 00:06:13,860
teach someone about programming hides a

167
00:06:11,940 --> 00:06:15,539
lot of complexity under the hood I I

168
00:06:13,860 --> 00:06:17,580
don't think that anyone would disagree

169
00:06:15,539 --> 00:06:19,620
that the best most maintainable code is

170
00:06:17,580 --> 00:06:21,240
the simplest code that you can write

171
00:06:19,620 --> 00:06:22,259
under the constraints that we have to

172
00:06:21,240 --> 00:06:24,780
operate in

173
00:06:22,259 --> 00:06:26,520
with software it exists because there is

174
00:06:24,780 --> 00:06:28,080
complexity inherent in the world and if

175
00:06:26,520 --> 00:06:30,240
we can model that complexity then then

176
00:06:28,080 --> 00:06:31,919
we help people navigate that complexity

177
00:06:30,240 --> 00:06:33,840
and by helping people navigate

178
00:06:31,919 --> 00:06:36,120
complexity if we do it right things

179
00:06:33,840 --> 00:06:38,580
become simpler for them now looking at

180
00:06:36,120 --> 00:06:40,860
print is is instructive it only appears

181
00:06:38,580 --> 00:06:43,080
uh simple whatever that is because we

182
00:06:40,860 --> 00:06:44,880
wrote building blocks and with each tool

183
00:06:43,080 --> 00:06:47,039
you're from the LCD elements in our

184
00:06:44,880 --> 00:06:49,560
projector here right up to the python

185
00:06:47,039 --> 00:06:52,319
interpreter itself more complexity gets

186
00:06:49,560 --> 00:06:54,000
modeled and simplified for us which is

187
00:06:52,319 --> 00:06:55,979
leading me to some indisputable facts

188
00:06:54,000 --> 00:06:58,680
here the first is that complexity exists

189
00:06:55,979 --> 00:07:00,600
that it will be experienced by somebody

190
00:06:58,680 --> 00:07:02,160
on a project of some description it

191
00:07:00,600 --> 00:07:04,199
could be an end user it could be you as

192
00:07:02,160 --> 00:07:06,419
the programmer and on a systems design

193
00:07:04,199 --> 00:07:07,740
level complexity is either modeled or

194
00:07:06,419 --> 00:07:10,139
ignored

195
00:07:07,740 --> 00:07:12,900
and more importantly that Simplicity

196
00:07:10,139 --> 00:07:15,660
usually belies some level of complexity

197
00:07:12,900 --> 00:07:17,639
on some level within a system

198
00:07:15,660 --> 00:07:19,440
put another way anyone who is

199
00:07:17,639 --> 00:07:21,360
experiencing Simplicity is doing so

200
00:07:19,440 --> 00:07:24,440
because complexity has been dealt with

201
00:07:21,360 --> 00:07:24,440
in some sort of way

202
00:07:24,780 --> 00:07:27,960
so with that in mind

203
00:07:26,460 --> 00:07:29,520
I used to think about how to build the

204
00:07:27,960 --> 00:07:32,039
simplest possible solution to a problem

205
00:07:29,520 --> 00:07:34,080
which which raises the question of what

206
00:07:32,039 --> 00:07:35,759
we might think is simple and I don't

207
00:07:34,080 --> 00:07:38,400
think this is a good question anymore I

208
00:07:35,759 --> 00:07:40,259
I don't think it scales well

209
00:07:38,400 --> 00:07:41,759
um instead if we accept that Simplicity

210
00:07:40,259 --> 00:07:43,919
will usually belie some sort of

211
00:07:41,759 --> 00:07:46,139
complexity within a system then the

212
00:07:43,919 --> 00:07:48,120
questions we should be asking is who

213
00:07:46,139 --> 00:07:49,199
should be experiencing Simplicity within

214
00:07:48,120 --> 00:07:51,360
a system

215
00:07:49,199 --> 00:07:52,440
which parts of our system should be

216
00:07:51,360 --> 00:07:54,240
simple

217
00:07:52,440 --> 00:07:55,919
and where should Simplicity be

218
00:07:54,240 --> 00:07:57,780
experienced

219
00:07:55,919 --> 00:07:59,220
and the main thing that I want to convey

220
00:07:57,780 --> 00:08:01,919
is that the answers to these questions

221
00:07:59,220 --> 00:08:04,620
are in flux python is an evolving

222
00:08:01,919 --> 00:08:06,419
ecosystem and as new problems arise and

223
00:08:04,620 --> 00:08:09,599
new tools make different problem domains

224
00:08:06,419 --> 00:08:13,520
easier to navigate our Focus changes and

225
00:08:09,599 --> 00:08:13,520
our Focus will continue to change

226
00:08:18,120 --> 00:08:23,340
um so the first idea that I want to

227
00:08:21,120 --> 00:08:25,860
explore is the idea of internalized

228
00:08:23,340 --> 00:08:27,720
versus externalized complexity which is

229
00:08:25,860 --> 00:08:30,900
one where python has really changed its

230
00:08:27,720 --> 00:08:32,880
tune over the last 10 13 15 years

231
00:08:30,900 --> 00:08:35,159
thereabouts there used to be this idea

232
00:08:32,880 --> 00:08:37,680
that python would come with everything

233
00:08:35,159 --> 00:08:39,180
that you needed to solve most problems

234
00:08:37,680 --> 00:08:40,919
you might ever encounter out of the box

235
00:08:39,180 --> 00:08:42,659
and this idea was prevalent back around

236
00:08:40,919 --> 00:08:44,159
the time the Zen of python was actually

237
00:08:42,659 --> 00:08:45,899
written

238
00:08:44,159 --> 00:08:47,580
is actually less relevant for these days

239
00:08:45,899 --> 00:08:49,560
for a number of reasons that's but not

240
00:08:47,580 --> 00:08:51,660
particularly important right now

241
00:08:49,560 --> 00:08:54,480
um who's seen this thing

242
00:08:51,660 --> 00:08:57,300
right firstly like you know ignore this

243
00:08:54,480 --> 00:08:58,920
a joke at the expense of pearl

244
00:08:57,300 --> 00:09:01,080
um this line was written as a joke

245
00:08:58,920 --> 00:09:04,019
there's still enough people uh in the

246
00:09:01,080 --> 00:09:06,360
python Community who take this seriously

247
00:09:04,019 --> 00:09:08,640
um and so python had this idea that if

248
00:09:06,360 --> 00:09:10,440
if it made the big decisions for us

249
00:09:08,640 --> 00:09:11,820
about how you'd solve your problems then

250
00:09:10,440 --> 00:09:13,260
everyone would have the same version of

251
00:09:11,820 --> 00:09:15,000
python and you could read another

252
00:09:13,260 --> 00:09:16,080
person's Python's code and that would be

253
00:09:15,000 --> 00:09:18,180
great and everybody could understand

254
00:09:16,080 --> 00:09:20,580
everyone's code so we're going to talk

255
00:09:18,180 --> 00:09:22,640
about a thing called URL lib2

256
00:09:20,580 --> 00:09:22,640
um

257
00:09:24,120 --> 00:09:28,140
so that that that two there that two

258
00:09:27,300 --> 00:09:30,779
there

259
00:09:28,140 --> 00:09:32,220
um that's entirely down to python not

260
00:09:30,779 --> 00:09:34,620
addressing the complexity of name

261
00:09:32,220 --> 00:09:36,600
spacing in the standard Library which

262
00:09:34,620 --> 00:09:38,640
they fixed in Python 3 they called it

263
00:09:36,600 --> 00:09:40,019
urlab dot request and and that

264
00:09:38,640 --> 00:09:41,640
introduced complexity and reading

265
00:09:40,019 --> 00:09:44,040
documentation which I'm not going to get

266
00:09:41,640 --> 00:09:46,320
into but that that's the thing as well

267
00:09:44,040 --> 00:09:47,820
um eurolib 2 was built around the idea

268
00:09:46,320 --> 00:09:50,220
that your code could be simpler if you

269
00:09:47,820 --> 00:09:51,839
could just drop URL handling code into

270
00:09:50,220 --> 00:09:54,540
python code that dealt with other

271
00:09:51,839 --> 00:09:56,279
interfaces and so it did all this in

272
00:09:54,540 --> 00:09:58,800
what would be a relatively familiar way

273
00:09:56,279 --> 00:10:00,779
to you uh your request objects they kind

274
00:09:58,800 --> 00:10:02,640
of Quack Like files there's an

275
00:10:00,779 --> 00:10:05,279
object-oriented interface for installing

276
00:10:02,640 --> 00:10:06,720
handlers for more complex use cases and

277
00:10:05,279 --> 00:10:08,399
you could install these handers for

278
00:10:06,720 --> 00:10:11,040
authentication and redirect handling

279
00:10:08,399 --> 00:10:11,880
either with or without affecting Global

280
00:10:11,040 --> 00:10:14,459
state

281
00:10:11,880 --> 00:10:15,720
and so in doing so it let you do a

282
00:10:14,459 --> 00:10:17,640
reasonable job of modeling the

283
00:10:15,720 --> 00:10:21,540
complexity of working with perfectly

284
00:10:17,640 --> 00:10:24,660
specified HTTP but how does that look in

285
00:10:21,540 --> 00:10:26,640
practice so URL Ed 2 has a thing called

286
00:10:24,660 --> 00:10:28,200
authentication helpers

287
00:10:26,640 --> 00:10:30,420
um what you do is you create a password

288
00:10:28,200 --> 00:10:33,300
manager to associate some credentials

289
00:10:30,420 --> 00:10:35,519
with specific URLs that's this this HTTP

290
00:10:33,300 --> 00:10:36,300
password blah blah blah blah blah blah

291
00:10:35,519 --> 00:10:37,200
thing

292
00:10:36,300 --> 00:10:39,120
um

293
00:10:37,200 --> 00:10:40,800
and you can create an authentication

294
00:10:39,120 --> 00:10:42,899
Handler that supplies these login

295
00:10:40,800 --> 00:10:45,120
credentials and then you create an

296
00:10:42,899 --> 00:10:47,040
opener that knows how to open URLs with

297
00:10:45,120 --> 00:10:48,839
these auth credentials and you use that

298
00:10:47,040 --> 00:10:50,940
opener to open the URL and then you

299
00:10:48,839 --> 00:10:53,100
treat the response like a file and you

300
00:10:50,940 --> 00:10:55,079
hand them into json.load which I presume

301
00:10:53,100 --> 00:10:57,860
you've you've imported somewhere in this

302
00:10:55,079 --> 00:10:57,860
in this code before

303
00:10:57,899 --> 00:11:03,480
this this code becomes visibly simpler

304
00:11:00,779 --> 00:11:05,459
if you choose for URL lib2 to not manage

305
00:11:03,480 --> 00:11:07,680
the complexity of authentication for you

306
00:11:05,459 --> 00:11:10,680
as long as you know how to write this

307
00:11:07,680 --> 00:11:13,260
this header you can make a request with

308
00:11:10,680 --> 00:11:14,820
that header and with this you're forced

309
00:11:13,260 --> 00:11:17,820
to make a trade-off

310
00:11:14,820 --> 00:11:20,700
you either get simple code and you model

311
00:11:17,820 --> 00:11:23,279
the complexity yourself or you have the

312
00:11:20,700 --> 00:11:25,320
complexity handled for you but you spend

313
00:11:23,279 --> 00:11:27,060
a lot of time rigging up this reusable

314
00:11:25,320 --> 00:11:29,459
authentication handling infrastructure

315
00:11:27,060 --> 00:11:32,880
and so the alternative is that we use a

316
00:11:29,459 --> 00:11:34,440
different Library like requests the

317
00:11:32,880 --> 00:11:36,540
behavior and authentication is a lot

318
00:11:34,440 --> 00:11:39,060
more declarative requests you say you

319
00:11:36,540 --> 00:11:40,860
want better off you make a request using

320
00:11:39,060 --> 00:11:42,300
those declarations and as a bonus you

321
00:11:40,860 --> 00:11:44,579
don't need to think about file handling

322
00:11:42,300 --> 00:11:46,440
because it Imports Json for you and this

323
00:11:44,579 --> 00:11:49,040
is obviously much simpler code to

324
00:11:46,440 --> 00:11:51,480
interact with an authenticated HTTP API

325
00:11:49,040 --> 00:11:52,800
you can make use of the libraries work

326
00:11:51,480 --> 00:11:55,019
to model the complexity of

327
00:11:52,800 --> 00:11:56,880
authentication and you get short code

328
00:11:55,019 --> 00:11:59,399
you don't need to make these trade-offs

329
00:11:56,880 --> 00:12:01,760
within your code but what's particularly

330
00:11:59,399 --> 00:12:04,320
interesting about this is that

331
00:12:01,760 --> 00:12:06,779
requests chose to re-implement something

332
00:12:04,320 --> 00:12:09,720
that existed in the standard library and

333
00:12:06,779 --> 00:12:11,820
then it choose it chose not to be in the

334
00:12:09,720 --> 00:12:13,800
standard Library itself

335
00:12:11,820 --> 00:12:16,800
this was a very odd choice at the time

336
00:12:13,800 --> 00:12:19,740
but in retrospect this is how things

337
00:12:16,800 --> 00:12:21,000
tend to get done in Python these days by

338
00:12:19,740 --> 00:12:23,220
being out of the standard Library

339
00:12:21,000 --> 00:12:25,560
request was able to adapt as a broader

340
00:12:23,220 --> 00:12:27,480
web communities changed their attitudes

341
00:12:25,560 --> 00:12:29,700
changed particularly around things like

342
00:12:27,480 --> 00:12:31,440
security and it shows to model some

343
00:12:29,700 --> 00:12:34,019
complexity of interacting with HTTP

344
00:12:31,440 --> 00:12:36,180
services for end users and as a result

345
00:12:34,019 --> 00:12:38,040
the code that you need to write to

346
00:12:36,180 --> 00:12:42,660
interact with requests is much much

347
00:12:38,040 --> 00:12:43,980
simpler than interacting with URL lib2

348
00:12:42,660 --> 00:12:46,260
um

349
00:12:43,980 --> 00:12:49,320
so the question that that arises from

350
00:12:46,260 --> 00:12:51,480
this is what do you get for that gaining

351
00:12:49,320 --> 00:12:53,760
Simplicity is this absolute

352
00:12:51,480 --> 00:12:55,380
and the answer is no this this gain is

353
00:12:53,760 --> 00:12:57,240
not absolute you now you need to model

354
00:12:55,380 --> 00:12:59,519
dependencies you need to have a

355
00:12:57,240 --> 00:13:01,100
requirements.txt you need to write

356
00:12:59,519 --> 00:13:03,300
instructions on on which

357
00:13:01,100 --> 00:13:05,100
requirements.txt file to load with Pip

358
00:13:03,300 --> 00:13:06,899
you need to manage execution

359
00:13:05,100 --> 00:13:08,940
environments you need to keep things up

360
00:13:06,899 --> 00:13:12,360
to date and so there's now added

361
00:13:08,940 --> 00:13:14,519
complexity in maintaining a project

362
00:13:12,360 --> 00:13:15,839
so which is the right choice here or if

363
00:13:14,519 --> 00:13:18,959
you're making the choice you use

364
00:13:15,839 --> 00:13:21,360
requests or urlab 2 in 2011 the choice

365
00:13:18,959 --> 00:13:22,620
might not have been immediately clear I

366
00:13:21,360 --> 00:13:24,300
think people were still thinking about

367
00:13:22,620 --> 00:13:26,399
whether using virtual amp was the right

368
00:13:24,300 --> 00:13:28,620
thing to do or not and this was still

369
00:13:26,399 --> 00:13:30,959
fairly early in the trend of HTTP apis

370
00:13:28,620 --> 00:13:33,899
where you know everything is an API now

371
00:13:30,959 --> 00:13:35,700
it wasn't the case back then so maybe if

372
00:13:33,899 --> 00:13:37,860
your application needed to hit a number

373
00:13:35,700 --> 00:13:39,540
of services that look like GitHub and

374
00:13:37,860 --> 00:13:41,459
regular you might do that

375
00:13:39,540 --> 00:13:42,899
I mean if you are fetching packages of a

376
00:13:41,459 --> 00:13:45,660
Debian mirror or something like that

377
00:13:42,899 --> 00:13:47,279
that doesn't require authentication you

378
00:13:45,660 --> 00:13:49,139
you are a lib2 might have been simpler

379
00:13:47,279 --> 00:13:50,579
for you and so you'd have you'd have a

380
00:13:49,139 --> 00:13:52,680
bit more first party code wherever you

381
00:13:50,579 --> 00:13:54,060
needed to hit a weird API but your

382
00:13:52,680 --> 00:13:55,740
overall code base would be smaller with

383
00:13:54,060 --> 00:13:57,959
urlib too and you wouldn't need to track

384
00:13:55,740 --> 00:13:59,820
external dependencies and the

385
00:13:57,959 --> 00:14:02,339
alternative would be to have less code

386
00:13:59,820 --> 00:14:04,800
but more complexity in terms of the

387
00:14:02,339 --> 00:14:07,200
environment that you need to manage

388
00:14:04,800 --> 00:14:08,940
I think in retrospect this idea of less

389
00:14:07,200 --> 00:14:11,040
first party code and letting your

390
00:14:08,940 --> 00:14:12,480
dependencies do more of the work is an

391
00:14:11,040 --> 00:14:14,279
approach that is one

392
00:14:12,480 --> 00:14:16,200
and this is not because absolutely

393
00:14:14,279 --> 00:14:18,300
everything was in place 10 years ago

394
00:14:16,200 --> 00:14:20,339
it's because we saw this as kind of a

395
00:14:18,300 --> 00:14:22,560
good way forward and we built tools to

396
00:14:20,339 --> 00:14:24,360
help us manage complexity that arose

397
00:14:22,560 --> 00:14:26,820
from deciding to manage external

398
00:14:24,360 --> 00:14:29,399
dependencies and configuration we have

399
00:14:26,820 --> 00:14:31,740
things like GitHub and and git more

400
00:14:29,399 --> 00:14:34,019
generally to manage our configuration

401
00:14:31,740 --> 00:14:36,240
management we have reasonable tooling

402
00:14:34,019 --> 00:14:39,240
for managing execution environments and

403
00:14:36,240 --> 00:14:40,800
we have a quite good package index

404
00:14:39,240 --> 00:14:42,300
and so now we're in a place where our

405
00:14:40,800 --> 00:14:44,760
definition of Simplicity has actually

406
00:14:42,300 --> 00:14:46,860
changed for an identical problem a

407
00:14:44,760 --> 00:14:48,720
modern python code base is going to be

408
00:14:46,860 --> 00:14:51,120
smaller than one that we wrote 10 to 15

409
00:14:48,720 --> 00:14:52,560
years ago to solve the same problem but

410
00:14:51,120 --> 00:14:55,019
the trade-off is that we have an

411
00:14:52,560 --> 00:14:57,360
increase in complexity of configuration

412
00:14:55,019 --> 00:15:00,240
management so we have more dependencies

413
00:14:57,360 --> 00:15:01,980
but we have more rigid environmental

414
00:15:00,240 --> 00:15:03,660
requirements

415
00:15:01,980 --> 00:15:06,000
but those dependencies they model

416
00:15:03,660 --> 00:15:07,980
complexity themselves now we have

417
00:15:06,000 --> 00:15:10,320
configuration tools they modal

418
00:15:07,980 --> 00:15:12,360
complexity and so for the modern python

419
00:15:10,320 --> 00:15:14,579
developer it seems that we get

420
00:15:12,360 --> 00:15:17,760
Simplicity as a virtue of other people

421
00:15:14,579 --> 00:15:19,019
managing complexity for us and just as a

422
00:15:17,760 --> 00:15:20,820
thought like if simple is better than

423
00:15:19,019 --> 00:15:22,680
complex well what does it say about how

424
00:15:20,820 --> 00:15:26,240
good the code that we're using to manage

425
00:15:22,680 --> 00:15:26,240
our environmental complexity is

426
00:15:27,199 --> 00:15:32,579
so the second idea I want to talk about

427
00:15:29,699 --> 00:15:34,680
is one that's less conclusively answered

428
00:15:32,579 --> 00:15:36,899
in Python which is the idea of how we go

429
00:15:34,680 --> 00:15:38,760
about managing complexity in a code base

430
00:15:36,899 --> 00:15:42,360
and whether the tools that we have

431
00:15:38,760 --> 00:15:44,820
actually present us uh present us with

432
00:15:42,360 --> 00:15:47,339
Simplicity and specifically I'm going to

433
00:15:44,820 --> 00:15:48,660
talk about quality assurance which is

434
00:15:47,339 --> 00:15:50,519
the set of tools that we use to make

435
00:15:48,660 --> 00:15:52,139
sure that our programs do what we say

436
00:15:50,519 --> 00:15:55,740
where we'll do what they're meant to do

437
00:15:52,139 --> 00:15:58,139
but first let's talk about duct typing

438
00:15:55,740 --> 00:15:59,639
so duct typing is an idea that two

439
00:15:58,139 --> 00:16:00,959
objects are interchangeable if you can

440
00:15:59,639 --> 00:16:03,360
use them in the same place and not get

441
00:16:00,959 --> 00:16:04,740
interface related errors if you're a

442
00:16:03,360 --> 00:16:06,899
type theorist you would call it

443
00:16:04,740 --> 00:16:07,920
structural polymorphism I'm not so I

444
00:16:06,899 --> 00:16:09,959
won't

445
00:16:07,920 --> 00:16:12,240
um Python's use of duct typing is a very

446
00:16:09,959 --> 00:16:14,579
very simple type system both to

447
00:16:12,240 --> 00:16:16,440
implement and to understand firstly

448
00:16:14,579 --> 00:16:17,940
there's no syntax in the language that

449
00:16:16,440 --> 00:16:19,620
describes the types that exist on an

450
00:16:17,940 --> 00:16:21,959
interface just have functions and

451
00:16:19,620 --> 00:16:23,880
variables this is simpler than Java or C

452
00:16:21,959 --> 00:16:26,399
plus plus which require you to enumerate

453
00:16:23,880 --> 00:16:27,839
the types that an interface expects and

454
00:16:26,399 --> 00:16:29,459
then on the other hand there's no

455
00:16:27,839 --> 00:16:32,459
automatic coercion so the language

456
00:16:29,459 --> 00:16:34,320
doesn't do complex Behavior to make data

457
00:16:32,459 --> 00:16:36,540
behave in different contexts this is

458
00:16:34,320 --> 00:16:39,000
simple as a JavaScript or Perl or PHP

459
00:16:36,540 --> 00:16:42,300
this forces some complexity in terms of

460
00:16:39,000 --> 00:16:45,540
managing types onto you as a developer

461
00:16:42,300 --> 00:16:47,399
um so when faced with some Library code

462
00:16:45,540 --> 00:16:48,660
you may ask yourself well how do you

463
00:16:47,399 --> 00:16:51,720
make sure that your library code is

464
00:16:48,660 --> 00:16:53,759
correct and that normally comes down to

465
00:16:51,720 --> 00:16:56,160
two questions which is what does your

466
00:16:53,759 --> 00:16:57,779
code do and where should your code be

467
00:16:56,160 --> 00:16:59,699
used

468
00:16:57,779 --> 00:17:01,620
so let's talk about this first one or

469
00:16:59,699 --> 00:17:03,480
how do we solve this what does your code

470
00:17:01,620 --> 00:17:05,699
do problem from in terms of quality

471
00:17:03,480 --> 00:17:08,339
assurance perspective

472
00:17:05,699 --> 00:17:10,199
um it's it's to do things you know test

473
00:17:08,339 --> 00:17:11,459
you know do your numeric inputs work do

474
00:17:10,199 --> 00:17:13,500
your strings work do your valid

475
00:17:11,459 --> 00:17:16,260
non-numeric inputs work

476
00:17:13,500 --> 00:17:18,120
things that you write with tests

477
00:17:16,260 --> 00:17:20,400
so automated tests help us manage

478
00:17:18,120 --> 00:17:22,079
complexity of verifying Behavior by

479
00:17:20,400 --> 00:17:23,339
allowing us to automatically run code

480
00:17:22,079 --> 00:17:26,520
that will check the code that we're

481
00:17:23,339 --> 00:17:27,839
verifying under the test works and to go

482
00:17:26,520 --> 00:17:29,520
back to this talking about computer

483
00:17:27,839 --> 00:17:30,840
science terminology stuff I was doing

484
00:17:29,520 --> 00:17:32,520
earlier

485
00:17:30,840 --> 00:17:34,799
um what we're doing is describing a set

486
00:17:32,520 --> 00:17:37,320
of things called invariance and

487
00:17:34,799 --> 00:17:39,419
invariant is basically a specification

488
00:17:37,320 --> 00:17:42,780
of some behavior that a system should

489
00:17:39,419 --> 00:17:44,640
hold given a set of circumstances if an

490
00:17:42,780 --> 00:17:46,740
invariant stops holding then either the

491
00:17:44,640 --> 00:17:48,600
behavior is wrong or the specification

492
00:17:46,740 --> 00:17:49,860
is wrong and the more invariants you

493
00:17:48,600 --> 00:17:52,320
describe the narrower the set of

494
00:17:49,860 --> 00:17:54,120
behaviors your system can display

495
00:17:52,320 --> 00:17:55,380
which means that the more invariants you

496
00:17:54,120 --> 00:17:58,020
have specified

497
00:17:55,380 --> 00:18:00,600
the more your complexity is managed

498
00:17:58,020 --> 00:18:03,179
and a unit test is a way to specify an

499
00:18:00,600 --> 00:18:04,740
invariant an invariant the circumstances

500
00:18:03,179 --> 00:18:06,960
of this data that you provide to the

501
00:18:04,740 --> 00:18:08,580
function under test and the invariant is

502
00:18:06,960 --> 00:18:10,380
a set of assertions that are specified

503
00:18:08,580 --> 00:18:12,360
about the result

504
00:18:10,380 --> 00:18:14,039
and this is really familiar to us it

505
00:18:12,360 --> 00:18:15,720
gives us a simple code-based structure

506
00:18:14,039 --> 00:18:16,919
to reason with which is that

507
00:18:15,720 --> 00:18:19,080
functionality lives in the

508
00:18:16,919 --> 00:18:22,559
implementation files you have and in

509
00:18:19,080 --> 00:18:24,240
variants are specified in test files

510
00:18:22,559 --> 00:18:25,860
but what if we think about our simple

511
00:18:24,240 --> 00:18:27,299
framework for quality insurance well

512
00:18:25,860 --> 00:18:29,280
we've only really looked at one of these

513
00:18:27,299 --> 00:18:32,220
two questions the question of what our

514
00:18:29,280 --> 00:18:33,780
code actually does if we're thinking

515
00:18:32,220 --> 00:18:35,940
about these and giving them fancy names

516
00:18:33,780 --> 00:18:37,860
we can classify these two questions into

517
00:18:35,940 --> 00:18:39,720
functional styling variants which are

518
00:18:37,860 --> 00:18:42,600
the things that specify how the thing

519
00:18:39,720 --> 00:18:44,580
works and structural ones which is how a

520
00:18:42,600 --> 00:18:47,340
piece of software fits into a system

521
00:18:44,580 --> 00:18:49,260
around it and the tool that we have for

522
00:18:47,340 --> 00:18:50,820
describing functional invariants I don't

523
00:18:49,260 --> 00:18:52,980
think are very good at describing

524
00:18:50,820 --> 00:18:54,960
structural invariance we can describe

525
00:18:52,980 --> 00:18:56,580
where things will work but we can't

526
00:18:54,960 --> 00:18:59,039
really exhaustively describe where

527
00:18:56,580 --> 00:19:00,360
things won't work and in Python for the

528
00:18:59,039 --> 00:19:03,559
longest time we kind of solve this

529
00:19:00,360 --> 00:19:03,559
problem with documentation

530
00:19:05,580 --> 00:19:09,900
on the other hand

531
00:19:07,200 --> 00:19:12,059
we have types now and types are great at

532
00:19:09,900 --> 00:19:14,280
specifying structural invariants and

533
00:19:12,059 --> 00:19:16,919
recent versions of python can describe

534
00:19:14,280 --> 00:19:18,600
types for most reasonable python code

535
00:19:16,919 --> 00:19:20,039
as well as providing structural and

536
00:19:18,600 --> 00:19:22,260
variance you also don't need to write as

537
00:19:20,039 --> 00:19:23,820
many tests so if you can't specify the

538
00:19:22,260 --> 00:19:25,080
wrong type you don't need to enumerate

539
00:19:23,820 --> 00:19:27,660
all the cases where you might have a

540
00:19:25,080 --> 00:19:29,220
type error you can't specify the wrong

541
00:19:27,660 --> 00:19:32,160
type so you don't need to test those

542
00:19:29,220 --> 00:19:34,200
things but I'm not here to evangelize

543
00:19:32,160 --> 00:19:35,760
types to you I think they're great but

544
00:19:34,200 --> 00:19:38,100
they still aren't universally popular

545
00:19:35,760 --> 00:19:41,160
and they're really really good reasons

546
00:19:38,100 --> 00:19:42,780
for that firstly types are a tool for

547
00:19:41,160 --> 00:19:46,020
managing complexity that really only

548
00:19:42,780 --> 00:19:47,760
occurs in large code bases and it's not

549
00:19:46,020 --> 00:19:49,559
always clear when you've crossed the

550
00:19:47,760 --> 00:19:52,020
line of structural complexity that

551
00:19:49,559 --> 00:19:54,419
arises where you can no longer fit

552
00:19:52,020 --> 00:19:56,220
things entirely within your head I I

553
00:19:54,419 --> 00:19:58,200
don't even think it's always clear the

554
00:19:56,220 --> 00:20:00,360
code base will ever achieve the level of

555
00:19:58,200 --> 00:20:02,220
structural complexity where you need to

556
00:20:00,360 --> 00:20:04,020
have typing in place

557
00:20:02,220 --> 00:20:05,940
and generally it can be hard to judge

558
00:20:04,020 --> 00:20:07,799
when types are necessary and once you do

559
00:20:05,940 --> 00:20:09,840
you kind of have to learn to use an

560
00:20:07,799 --> 00:20:11,520
entirely new system which itself can be

561
00:20:09,840 --> 00:20:13,440
a complex thing

562
00:20:11,520 --> 00:20:14,940
and so you end up with a trade-off you

563
00:20:13,440 --> 00:20:16,980
can either have a simpler toolbox for

564
00:20:14,940 --> 00:20:18,960
managing quality assurance or you can

565
00:20:16,980 --> 00:20:20,640
have a simpler way to specify the

566
00:20:18,960 --> 00:20:22,200
structure of your code base

567
00:20:20,640 --> 00:20:24,299
and so you can either have more tests

568
00:20:22,200 --> 00:20:25,980
and more documentation or you can make

569
00:20:24,299 --> 00:20:27,660
your interfaces harder to read and

570
00:20:25,980 --> 00:20:29,760
harder to understand

571
00:20:27,660 --> 00:20:31,380
and so there isn't actually a correct

572
00:20:29,760 --> 00:20:33,419
answer here yet I don't think I think

573
00:20:31,380 --> 00:20:35,400
both approaches provide some sort of

574
00:20:33,419 --> 00:20:37,620
Simplicity depending on what sort of

575
00:20:35,400 --> 00:20:39,539
problem you're trying to solve but they

576
00:20:37,620 --> 00:20:42,900
both leave complexity elsewhere in a

577
00:20:39,539 --> 00:20:44,460
project depending on how big it gets

578
00:20:42,900 --> 00:20:46,620
so I'd like to finish up with a

579
00:20:44,460 --> 00:20:48,660
discussion of evolution of python itself

580
00:20:46,620 --> 00:20:51,059
because a common complaint I hear is

581
00:20:48,660 --> 00:20:53,640
that that python is becoming more

582
00:20:51,059 --> 00:20:55,559
complex with every single version and I

583
00:20:53,640 --> 00:20:59,179
think it's worth taking some time to

584
00:20:55,559 --> 00:20:59,179
examine whether that is the case

585
00:21:02,460 --> 00:21:05,460
laughs

586
00:21:09,179 --> 00:21:14,820
so python was for a long time a

587
00:21:13,080 --> 00:21:18,059
completely static language for a lot of

588
00:21:14,820 --> 00:21:19,919
us Python 2.7 was around for a very long

589
00:21:18,059 --> 00:21:21,660
time and that's kind of the time when a

590
00:21:19,919 --> 00:21:23,160
lot of the growth of of python actually

591
00:21:21,660 --> 00:21:25,559
happened and was only really around

592
00:21:23,160 --> 00:21:27,900
python 3.5 that the language started to

593
00:21:25,559 --> 00:21:30,360
become stable for people who needed to

594
00:21:27,900 --> 00:21:32,580
do python who needed to use Python 3 for

595
00:21:30,360 --> 00:21:33,900
things other than data science and the

596
00:21:32,580 --> 00:21:35,760
language is kind of been adding on

597
00:21:33,900 --> 00:21:38,820
language features very regularly with

598
00:21:35,760 --> 00:21:40,740
every single major release of python

599
00:21:38,820 --> 00:21:42,659
um but if you're used to python not

600
00:21:40,740 --> 00:21:44,880
changing on you for seven whole years

601
00:21:42,659 --> 00:21:46,620
then maybe you'd be asking yourself well

602
00:21:44,880 --> 00:21:49,140
what happened to this idea of of simple

603
00:21:46,620 --> 00:21:54,020
being better than complex

604
00:21:49,140 --> 00:21:54,020
um to which I would respond with print

605
00:21:54,299 --> 00:21:57,780
because python has always been a

606
00:21:55,919 --> 00:21:59,940
language that kind of tries to model

607
00:21:57,780 --> 00:22:02,460
complexity so that users can write

608
00:21:59,940 --> 00:22:05,460
simpler code and one thing that happens

609
00:22:02,460 --> 00:22:07,919
as more people from more diverse use

610
00:22:05,460 --> 00:22:10,140
cases decide to use Python to solve

611
00:22:07,919 --> 00:22:12,840
their problems is that we discover more

612
00:22:10,140 --> 00:22:13,860
cases where python is just not simple to

613
00:22:12,840 --> 00:22:16,620
use yet

614
00:22:13,860 --> 00:22:18,299
so python just models more complexity

615
00:22:16,620 --> 00:22:21,059
for us

616
00:22:18,299 --> 00:22:23,460
most of us really when we first learned

617
00:22:21,059 --> 00:22:25,980
python only really picked up a subset of

618
00:22:23,460 --> 00:22:28,799
python as a language and likewise as new

619
00:22:25,980 --> 00:22:31,080
features emerge we'll only ever need a

620
00:22:28,799 --> 00:22:33,419
new a subset of these new features as

621
00:22:31,080 --> 00:22:35,520
somebody who who writes apis for things

622
00:22:33,419 --> 00:22:37,020
I don't need a matrix multiplication

623
00:22:35,520 --> 00:22:39,059
operator

624
00:22:37,020 --> 00:22:40,440
but then again a data analyst is

625
00:22:39,059 --> 00:22:42,480
probably never going to need to write an

626
00:22:40,440 --> 00:22:44,280
async await Co routine in their Jupiter

627
00:22:42,480 --> 00:22:47,220
notebook

628
00:22:44,280 --> 00:22:48,780
um I was at kiwi pycon last year and I

629
00:22:47,220 --> 00:22:50,640
gave a talk there and I tried to make a

630
00:22:48,780 --> 00:22:52,919
point that Python's success was its

631
00:22:50,640 --> 00:22:55,200
willingness to support use cases from a

632
00:22:52,919 --> 00:22:57,539
diverse set of disciplines another talk

633
00:22:55,200 --> 00:22:59,580
called the when of python argued that

634
00:22:57,539 --> 00:23:01,500
you can very easily use a subset of

635
00:22:59,580 --> 00:23:03,059
Python's features and never need to

636
00:23:01,500 --> 00:23:05,880
think about the more complex features

637
00:23:03,059 --> 00:23:08,340
until you actually need to use them and

638
00:23:05,880 --> 00:23:10,320
they even proposed a model for it it's

639
00:23:08,340 --> 00:23:11,760
called The when of python I don't

640
00:23:10,320 --> 00:23:13,260
entirely agree with it but I think it's

641
00:23:11,760 --> 00:23:15,500
a good start and you should look it up

642
00:23:13,260 --> 00:23:19,860
and either adopt it or criticize it

643
00:23:15,500 --> 00:23:21,919
basically go to last year's kiwi pycon

644
00:23:19,860 --> 00:23:23,880
uh there are videos

645
00:23:21,919 --> 00:23:25,860
that's the secret like this this

646
00:23:23,880 --> 00:23:28,140
complexity is reasonably self-contained

647
00:23:25,860 --> 00:23:30,900
and it's optional until you need it and

648
00:23:28,140 --> 00:23:33,539
this stands in contrast to needlessly

649
00:23:30,900 --> 00:23:37,980
complex languages like Java or even

650
00:23:33,539 --> 00:23:41,100
necessarily complex languages like rust

651
00:23:37,980 --> 00:23:42,840
all right we we as programmers want to

652
00:23:41,100 --> 00:23:44,640
continue to model more things and then

653
00:23:42,840 --> 00:23:46,260
the question we have to ask ourselves is

654
00:23:44,640 --> 00:23:48,780
well do we want to have a simple

655
00:23:46,260 --> 00:23:50,400
language or do we want to write simpler

656
00:23:48,780 --> 00:23:52,880
code with the language that we end up

657
00:23:50,400 --> 00:23:52,880
using

658
00:23:53,159 --> 00:23:57,179
um right so now I ended at the end of

659
00:23:55,020 --> 00:23:58,679
the examples section of this talk and

660
00:23:57,179 --> 00:24:00,299
we're going to enter into the realm of

661
00:23:58,679 --> 00:24:02,220
commentary again

662
00:24:00,299 --> 00:24:04,620
um so when we started

663
00:24:02,220 --> 00:24:06,419
we had this proposition that that simple

664
00:24:04,620 --> 00:24:08,100
is better than complex and we went

665
00:24:06,419 --> 00:24:09,720
through a bunch of examples that show

666
00:24:08,100 --> 00:24:11,580
that when you are writing python

667
00:24:09,720 --> 00:24:14,400
somebody is going to have to model

668
00:24:11,580 --> 00:24:15,780
complexity for us and in the world of

669
00:24:14,400 --> 00:24:17,700
python we're really lucky that the

670
00:24:15,780 --> 00:24:19,799
language we use chooses to Model A lot

671
00:24:17,700 --> 00:24:22,320
of complexity for us and it provides

672
00:24:19,799 --> 00:24:24,240
tools that can model new and unexpected

673
00:24:22,320 --> 00:24:25,799
forms of complexity

674
00:24:24,240 --> 00:24:28,140
but I haven't really spoken about code

675
00:24:25,799 --> 00:24:30,179
that that we write ourselves which

676
00:24:28,140 --> 00:24:31,380
brings me to this thing that I said at

677
00:24:30,179 --> 00:24:33,299
the top of the talk which is that

678
00:24:31,380 --> 00:24:35,100
complexity exists

679
00:24:33,299 --> 00:24:38,159
it would be better if complexity didn't

680
00:24:35,100 --> 00:24:40,799
exist but it does and

681
00:24:38,159 --> 00:24:42,960
as developers we can choose to contain

682
00:24:40,799 --> 00:24:45,900
that complexity or we can choose to do

683
00:24:42,960 --> 00:24:47,820
not contain that complexity

684
00:24:45,900 --> 00:24:49,860
so what I hope I've conveyed today is

685
00:24:47,820 --> 00:24:51,659
that we should be thinking about who

686
00:24:49,860 --> 00:24:54,179
should be experiencing Simplicity and

687
00:24:51,659 --> 00:24:55,679
who should experience complexity you

688
00:24:54,179 --> 00:24:57,780
know we work in narrow domains as

689
00:24:55,679 --> 00:24:59,640
developers and so even as if we're

690
00:24:57,780 --> 00:25:01,799
working in a team on the same project

691
00:24:59,640 --> 00:25:03,380
different people focus on different

692
00:25:01,799 --> 00:25:05,400
things

693
00:25:03,380 --> 00:25:07,020
with each of us having a pretty good

694
00:25:05,400 --> 00:25:08,340
understanding of of the stuff that we

695
00:25:07,020 --> 00:25:10,080
work on ourselves and so it's probably

696
00:25:08,340 --> 00:25:12,299
best if we mod our own domains

697
00:25:10,080 --> 00:25:13,620
complexity so it's not complicated for

698
00:25:12,299 --> 00:25:15,960
other people

699
00:25:13,620 --> 00:25:19,320
and this is why you have a giant pile of

700
00:25:15,960 --> 00:25:21,900
C code that does a whole bunch of stuff

701
00:25:19,320 --> 00:25:24,299
and then present a coherent interface

702
00:25:21,900 --> 00:25:25,799
that is safe to give to new programmers

703
00:25:24,299 --> 00:25:27,600
is literally the first thing we teach

704
00:25:25,799 --> 00:25:29,640
them

705
00:25:27,600 --> 00:25:32,400
like that's the point our code has end

706
00:25:29,640 --> 00:25:34,860
users whether they're consumers or

707
00:25:32,400 --> 00:25:37,559
whether their other software developers

708
00:25:34,860 --> 00:25:39,480
on your own team your code is almost

709
00:25:37,559 --> 00:25:41,760
always going to be part of some bigger

710
00:25:39,480 --> 00:25:44,039
system and it will exist because it's

711
00:25:41,760 --> 00:25:46,500
modeling some sort of complexity better

712
00:25:44,039 --> 00:25:48,720
than a human could your code has

713
00:25:46,500 --> 00:25:50,640
dependencies each of these dependencies

714
00:25:48,720 --> 00:25:52,500
from the python interpreter to the

715
00:25:50,640 --> 00:25:55,080
packages your dependencies depend upon

716
00:25:52,500 --> 00:25:57,720
model some sort of complexity so you

717
00:25:55,080 --> 00:25:59,100
don't have to otherwise it just wouldn't

718
00:25:57,720 --> 00:26:01,620
be useful for you and it wouldn't need

719
00:25:59,100 --> 00:26:04,320
to exist and like maybe your own code is

720
00:26:01,620 --> 00:26:06,659
somebody else's dependency chain

721
00:26:04,320 --> 00:26:08,880
we've seen earlier in this talk that the

722
00:26:06,659 --> 00:26:10,559
things that may be simplest they can

723
00:26:08,880 --> 00:26:12,720
change as we get newer tooling that

724
00:26:10,559 --> 00:26:14,220
solves problems more repeatedly or if

725
00:26:12,720 --> 00:26:15,240
people's opinions about what is good

726
00:26:14,220 --> 00:26:17,100
changes

727
00:26:15,240 --> 00:26:19,140
we've seen that what's simplest may

728
00:26:17,100 --> 00:26:21,539
depend on a trade-off between more

729
00:26:19,140 --> 00:26:24,659
upfront work and deferring decisions

730
00:26:21,539 --> 00:26:26,400
until later I for me a guiding principle

731
00:26:24,659 --> 00:26:29,520
is that Simplicity can be a gift that

732
00:26:26,400 --> 00:26:31,200
you give to others if if simple is

733
00:26:29,520 --> 00:26:33,779
really better than complex then leaving

734
00:26:31,200 --> 00:26:35,580
your end users with a simpler experience

735
00:26:33,779 --> 00:26:37,320
in managing your own complexity is

736
00:26:35,580 --> 00:26:39,059
probably better because that means that

737
00:26:37,320 --> 00:26:41,520
you're the only person dealing with this

738
00:26:39,059 --> 00:26:44,419
dealing with complexity that you have to

739
00:26:41,520 --> 00:26:46,860
deal with also I lied

740
00:26:44,419 --> 00:26:49,860
this is a slide that suggests that you

741
00:26:46,860 --> 00:26:52,080
should definitely use types if only at

742
00:26:49,860 --> 00:26:54,179
your interface typeset your interface

743
00:26:52,080 --> 00:26:56,220
give your users less complexity to need

744
00:26:54,179 --> 00:26:58,200
to worry about because you've told them

745
00:26:56,220 --> 00:27:00,000
what things will never work with your

746
00:26:58,200 --> 00:27:01,919
code

747
00:27:00,000 --> 00:27:03,960
choose to model complexly for your end

748
00:27:01,919 --> 00:27:05,779
users and annotate your interfaces with

749
00:27:03,960 --> 00:27:08,220
types if nothing else

750
00:27:05,779 --> 00:27:09,960
more generally like simplify things

751
00:27:08,220 --> 00:27:11,400
appropriately because I think this is

752
00:27:09,960 --> 00:27:13,500
what you should take away from this talk

753
00:27:11,400 --> 00:27:14,640
the idea that simple interfaces are a

754
00:27:13,500 --> 00:27:16,740
thing that we should strive for and

755
00:27:14,640 --> 00:27:19,880
complex ones are not but if you simplify

756
00:27:16,740 --> 00:27:19,880
things in the wrong place

757
00:27:21,440 --> 00:27:26,480
you tend to remove so much complexity

758
00:27:23,940 --> 00:27:30,860
for us to be useless and inactionable

759
00:27:26,480 --> 00:27:30,860
and you get talks like this one

760
00:27:32,100 --> 00:27:39,769
foreign

761
00:27:33,260 --> 00:27:39,769
[Applause]

762
00:27:39,960 --> 00:27:43,740
okay apparently we have time for

763
00:27:41,760 --> 00:27:46,919
questions so jacinta is going to hand a

764
00:27:43,740 --> 00:27:50,000
microphone somewhere and we'll see if

765
00:27:46,919 --> 00:27:50,000
anybody has a question

766
00:27:54,240 --> 00:27:57,980
you can ask the obvious one as well

767
00:27:58,980 --> 00:28:01,700
yes Tom

768
00:28:02,700 --> 00:28:05,900
what's the obvious question

769
00:28:06,480 --> 00:28:10,400
uh do you have your laptop out yeah you

770
00:28:08,400 --> 00:28:13,559
just type import this

771
00:28:10,400 --> 00:28:16,140
into a python interpreter okay yeah and

772
00:28:13,559 --> 00:28:17,039
just read what it says there to yourself

773
00:28:16,140 --> 00:28:20,240
um

774
00:28:17,039 --> 00:28:20,240
I'll just give you a moment

775
00:28:20,400 --> 00:28:24,720
have you have you found it

776
00:28:23,100 --> 00:28:27,260
okay well do you want to ask what the

777
00:28:24,720 --> 00:28:27,260
question is then

778
00:28:30,360 --> 00:28:35,640
well I guess now the question is um

779
00:28:33,840 --> 00:28:38,000
how do you know if it's simple if you're

780
00:28:35,640 --> 00:28:38,000
Dutch

781
00:28:38,940 --> 00:28:42,000
and that wasn't the question I'm sorry

782
00:28:40,500 --> 00:28:45,080
does anyone else know what the question

783
00:28:42,000 --> 00:28:45,080
I've prepared for is

784
00:28:45,120 --> 00:28:47,299
foreign

785
00:28:48,840 --> 00:28:55,940
what is the question you prepared for

786
00:28:52,080 --> 00:28:55,940
no no you don't get that sorry

787
00:29:02,179 --> 00:29:05,419
thanks everyone

788
00:29:06,059 --> 00:29:10,159
um is complex better than complicated

789
00:29:13,799 --> 00:29:17,940
I'm not answering that

790
00:29:15,130 --> 00:29:19,260
[Laughter]

791
00:29:17,940 --> 00:29:20,670
all right thanks everyone for coming

792
00:29:19,260 --> 00:29:27,150
along

793
00:29:20,670 --> 00:29:27,150
[Applause]