1
00:00:04,960 --> 00:00:19,999
[Music]

2
00:00:20,560 --> 00:00:25,599
I love that intro welcome back to the

3
00:00:22,760 --> 00:00:28,160
second talk of shiny block uh on Deck we

4
00:00:25,599 --> 00:00:32,480
have no CTS with what python can learn

5
00:00:28,160 --> 00:00:32,480
from other languages please welcome

6
00:00:36,320 --> 00:00:42,120
Noah see excellent I am indeed Noz I'm

7
00:00:39,960 --> 00:00:43,640
an S geom magical Labs we do computer

8
00:00:42,120 --> 00:00:45,440
vision and augmented reality stuff for

9
00:00:43,640 --> 00:00:47,360
Ikea but I'm not here to talk about that

10
00:00:45,440 --> 00:00:49,559
today and I'm also not really here to

11
00:00:47,360 --> 00:00:51,280
talk about python python is great it's

12
00:00:49,559 --> 00:00:52,879
my primary tool for solving most

13
00:00:51,280 --> 00:00:54,359
technical challenges but there's a lot

14
00:00:52,879 --> 00:00:55,800
of other programming languages out there

15
00:00:54,359 --> 00:00:57,800
and some of them have features or

16
00:00:55,800 --> 00:01:00,559
workflows that python

17
00:00:57,800 --> 00:01:02,039
lacks the goal of this talk is not to

18
00:01:00,559 --> 00:01:03,920
immediately Advocate that we turn each

19
00:01:02,039 --> 00:01:05,519
of these features into a pep we want to

20
00:01:03,920 --> 00:01:07,000
broaden our exposure to programming

21
00:01:05,519 --> 00:01:09,080
Concepts and talk about the trade-offs

22
00:01:07,000 --> 00:01:10,680
being made if any of these ideas really

23
00:01:09,080 --> 00:01:13,080
grab you try messing around with them in

24
00:01:10,680 --> 00:01:15,360
a library or make a patch or cpython but

25
00:01:13,080 --> 00:01:17,200
really we're just here to learn I'm

26
00:01:15,360 --> 00:01:18,720
going to begin with PHP and I don't mean

27
00:01:17,200 --> 00:01:21,360
this ironically or to be some kind of

28
00:01:18,720 --> 00:01:22,720
language hipster PHP is pretty cool I

29
00:01:21,360 --> 00:01:24,720
spent the last few years on a side

30
00:01:22,720 --> 00:01:26,200
project almost entirely in PHP and it's

31
00:01:24,720 --> 00:01:27,840
given me a refreshed

32
00:01:26,200 --> 00:01:30,119
appreciation the biggest thing I've

33
00:01:27,840 --> 00:01:32,280
grown to like about it the simplicity

34
00:01:30,119 --> 00:01:34,280
each file maps unambiguously to a single

35
00:01:32,280 --> 00:01:36,520
URL and deployment is as simple as

36
00:01:34,280 --> 00:01:38,159
updating that file in any way you prefer

37
00:01:36,520 --> 00:01:40,680
most editors these days support remote

38
00:01:38,159 --> 00:01:43,439
editing via SFTP so you just click on

39
00:01:40,680 --> 00:01:45,119
new file you start coding and that's it

40
00:01:43,439 --> 00:01:47,719
this puts even low boilerplate web

41
00:01:45,119 --> 00:01:49,240
Frameworks in Python to Absolute shame

42
00:01:47,719 --> 00:01:50,680
and once you get going changes in the

43
00:01:49,240 --> 00:01:52,560
code they just show up you don't need to

44
00:01:50,680 --> 00:01:54,960
restart a server or configure automatic

45
00:01:52,560 --> 00:01:58,200
reloads or whatever you just write code

46
00:01:54,960 --> 00:01:59,560
and it runs simple as that in Python we

47
00:01:58,200 --> 00:02:02,240
like to say that explicit is better than

48
00:01:59,560 --> 00:02:04,399
implicit it and overall I agree but man

49
00:02:02,240 --> 00:02:05,960
our juggling Imports annoying sometimes

50
00:02:04,399 --> 00:02:07,439
even with editor help you've got a dance

51
00:02:05,960 --> 00:02:08,560
of you write a new line of code you

52
00:02:07,439 --> 00:02:09,800
realize you don't have the import in

53
00:02:08,560 --> 00:02:11,239
place you scroll up to the top of the

54
00:02:09,800 --> 00:02:12,440
file you add the import you scroll back

55
00:02:11,239 --> 00:02:14,959
down to where you were you've lost your

56
00:02:12,440 --> 00:02:16,640
train of thought it's really jarring

57
00:02:14,959 --> 00:02:19,000
autoloaders bridge this Gap while

58
00:02:16,640 --> 00:02:21,480
keeping things mostly explicit if you

59
00:02:19,000 --> 00:02:23,400
enforce a onetoone mapping between uh

60
00:02:21,480 --> 00:02:24,959
code symbols and file system paths you

61
00:02:23,400 --> 00:02:26,840
know exactly where a symbol load is

62
00:02:24,959 --> 00:02:28,519
going to end up this is also deeply

63
00:02:26,840 --> 00:02:30,720
integrated into the PHP package manager

64
00:02:28,519 --> 00:02:31,920
called composer toe bake the mapping

65
00:02:30,720 --> 00:02:33,440
between those code symbols and the

66
00:02:31,920 --> 00:02:35,160
underlying file system paths so the

67
00:02:33,440 --> 00:02:37,160
import doesn't even have to Traverse a

68
00:02:35,160 --> 00:02:38,440
package directory I don't actually know

69
00:02:37,160 --> 00:02:39,640
if I would honestly prefer this in

70
00:02:38,440 --> 00:02:40,680
Python but it's at least something to

71
00:02:39,640 --> 00:02:43,360
think

72
00:02:40,680 --> 00:02:44,599
about the standard library for PHP is a

73
00:02:43,360 --> 00:02:46,400
little bit different from python it's

74
00:02:44,599 --> 00:02:48,319
evolved towards a single Niche web

75
00:02:46,400 --> 00:02:49,959
application development so it biases

76
00:02:48,319 --> 00:02:51,959
towards things needed in that space so

77
00:02:49,959 --> 00:02:53,920
for example my SQL and postgress clients

78
00:02:51,959 --> 00:02:56,120
come by default there's also a high

79
00:02:53,920 --> 00:02:57,560
performance HTTP client and then beyond

80
00:02:56,120 --> 00:03:00,360
that there's this thing called the PHP

81
00:02:57,560 --> 00:03:02,040
extension community library or Pele an

82
00:03:00,360 --> 00:03:03,879
extension PHP is very similar to an

83
00:03:02,040 --> 00:03:06,159
extension library in Python but the big

84
00:03:03,879 --> 00:03:08,799
difference between PE and pii is that

85
00:03:06,159 --> 00:03:11,120
for the a lot of the big core supported

86
00:03:08,799 --> 00:03:12,920
extensions the documentation for them is

87
00:03:11,120 --> 00:03:14,280
Right alongside the main PHP docs you

88
00:03:12,920 --> 00:03:16,040
don't have to search for a library and

89
00:03:14,280 --> 00:03:18,519
then find its read me or read the docs

90
00:03:16,040 --> 00:03:20,560
or whatever it's all just there Peck

91
00:03:18,519 --> 00:03:22,720
effectively bridges into a broader

92
00:03:20,560 --> 00:03:24,680
secondary standard

93
00:03:22,720 --> 00:03:26,440
library and alongside the standard

94
00:03:24,680 --> 00:03:29,560
Library being adapted for its major use

95
00:03:26,440 --> 00:03:31,120
case PHP as a language is itself too

96
00:03:29,560 --> 00:03:33,000
it's web first so there's a simple

97
00:03:31,120 --> 00:03:35,280
templating framework baked directly into

98
00:03:33,000 --> 00:03:36,760
the language in Python we have we have F

99
00:03:35,280 --> 00:03:38,239
strings now and they've matured over

100
00:03:36,760 --> 00:03:40,280
time we've definitely started moving in

101
00:03:38,239 --> 00:03:43,280
this direction but relatively complex

102
00:03:40,280 --> 00:03:46,560
PHP code is again going to run circles

103
00:03:43,280 --> 00:03:48,720
around Python's explicitly simplest web

104
00:03:46,560 --> 00:03:50,560
framworks Ruby is probably the second

105
00:03:48,720 --> 00:03:52,239
most language used over my career and

106
00:03:50,560 --> 00:03:53,480
while the days of python and Ruby being

107
00:03:52,239 --> 00:03:55,239
thought of as relatively neck and- neck

108
00:03:53,480 --> 00:03:56,840
have wed as Python's taken over more and

109
00:03:55,239 --> 00:03:58,159
more niches it's still a vibrant

110
00:03:56,840 --> 00:04:00,319
language in

111
00:03:58,159 --> 00:04:01,400
community symbols are thing in a lot of

112
00:04:00,319 --> 00:04:03,439
languages but I'm going to highlight

113
00:04:01,400 --> 00:04:05,680
them here at their simplest symbols are

114
00:04:03,439 --> 00:04:07,439
just fancy strings but internally

115
00:04:05,680 --> 00:04:09,040
they're D duplicated so checking

116
00:04:07,439 --> 00:04:11,079
equality between two symbols is a

117
00:04:09,040 --> 00:04:13,760
constant time operation we do have this

118
00:04:11,079 --> 00:04:16,040
in Python cy. intern and intern strings

119
00:04:13,760 --> 00:04:17,759
but they're very rarely used but there's

120
00:04:16,040 --> 00:04:19,959
this deeper level to symbols having a

121
00:04:17,759 --> 00:04:21,519
dedicated syntax for this makes it e

122
00:04:19,959 --> 00:04:23,960
easier to differentiate between text

123
00:04:21,519 --> 00:04:26,759
strings and for lack of a better term

124
00:04:23,960 --> 00:04:28,240
symbolic strings these often serve very

125
00:04:26,759 --> 00:04:29,600
different functions in code and even if

126
00:04:28,240 --> 00:04:31,919
they're the same underlying data type

127
00:04:29,600 --> 00:04:33,160
they they really shouldn't be mixed we

128
00:04:31,919 --> 00:04:35,160
do already kind of have this a little

129
00:04:33,160 --> 00:04:37,199
bit in Python with the uh keyword

130
00:04:35,160 --> 00:04:39,039
argument syntax those are still strings

131
00:04:37,199 --> 00:04:40,479
under the hood but they're clearly

132
00:04:39,039 --> 00:04:42,320
intended to be used for a different

133
00:04:40,479 --> 00:04:44,560
purpose and you don't uh you create a

134
00:04:42,320 --> 00:04:45,960
boundary between what kinds of use cases

135
00:04:44,560 --> 00:04:47,960
you want those strings to be

136
00:04:45,960 --> 00:04:50,560
used blocks are something much more

137
00:04:47,960 --> 00:04:52,400
specific to Ruby deep down we do have

138
00:04:50,560 --> 00:04:54,000
this in Python you can write a Lambda

139
00:04:52,400 --> 00:04:56,600
expression as an anonymous function

140
00:04:54,000 --> 00:04:58,919
value uh but blocks do a lot more than

141
00:04:56,600 --> 00:05:00,400
that first off they're not limited to a

142
00:04:58,919 --> 00:05:01,880
single expression that's handy for

143
00:05:00,400 --> 00:05:04,440
anything complicated but more

144
00:05:01,880 --> 00:05:06,560
importantly that calling syntax with do

145
00:05:04,440 --> 00:05:09,440
you can hang a block off any function in

146
00:05:06,560 --> 00:05:10,960
Ruby now you can pass around Lambda

147
00:05:09,440 --> 00:05:12,560
literals in Python it's technically

148
00:05:10,960 --> 00:05:14,600
possible we all sometimes do it when

149
00:05:12,560 --> 00:05:16,759
we're writing sort keys but the utility

150
00:05:14,600 --> 00:05:18,160
falls off really quickly after that and

151
00:05:16,759 --> 00:05:20,000
so Community Practice is that you'll

152
00:05:18,160 --> 00:05:22,319
write a named inline function and then

153
00:05:20,000 --> 00:05:24,080
you'll pass it around as a value totally

154
00:05:22,319 --> 00:05:25,960
works but it does have this problem that

155
00:05:24,080 --> 00:05:27,240
the read order is Now inverted from the

156
00:05:25,960 --> 00:05:29,360
execution

157
00:05:27,240 --> 00:05:30,800
order and this gets even worse when we

158
00:05:29,360 --> 00:05:32,720
start talking about method chaining

159
00:05:30,800 --> 00:05:34,319
something Ruby really excels at I'm not

160
00:05:32,720 --> 00:05:35,919
even really going to touch the python

161
00:05:34,319 --> 00:05:37,520
buil-in functions which are completely

162
00:05:35,919 --> 00:05:39,840
backwards from this and I hate it but

163
00:05:37,520 --> 00:05:42,160
you can write a method chaining API in

164
00:05:39,840 --> 00:05:43,800
Python for instance the Jenga omm but

165
00:05:42,160 --> 00:05:45,520
blocks allow it to go a lot further

166
00:05:43,800 --> 00:05:47,520
without additional

167
00:05:45,520 --> 00:05:49,199
effort and another language that has a

168
00:05:47,520 --> 00:05:51,720
different take on this is Dart and it's

169
00:05:49,199 --> 00:05:53,880
double do cascading operator this allows

170
00:05:51,720 --> 00:05:55,440
interrupting a method chain to go one

171
00:05:53,880 --> 00:05:57,400
level up so to speak like in a file

172
00:05:55,440 --> 00:05:59,400
system path performing a sequence of

173
00:05:57,400 --> 00:06:01,319
operations on the same object rather

174
00:05:59,400 --> 00:06:03,199
than to propagate the chain through

175
00:06:01,319 --> 00:06:05,479
every Link in it this allows method

176
00:06:03,199 --> 00:06:06,960
chaining in Dart with zero extra work by

177
00:06:05,479 --> 00:06:08,880
API

178
00:06:06,960 --> 00:06:10,319
authors another place really makes

179
00:06:08,880 --> 00:06:12,479
excellent use of blocks is its version

180
00:06:10,319 --> 00:06:14,120
of the Builder pattern many complex

181
00:06:12,479 --> 00:06:15,919
object instructors take a block which

182
00:06:14,120 --> 00:06:18,560
exposes a little configuration API

183
00:06:15,919 --> 00:06:21,160
almost a DSL this can allow for very

184
00:06:18,560 --> 00:06:22,800
complex initialization uh while still

185
00:06:21,160 --> 00:06:24,479
being Compact and without sacrificing

186
00:06:22,800 --> 00:06:26,039
flexibility for an usual conditions and

187
00:06:24,479 --> 00:06:27,639
it's very nice for things like static

188
00:06:26,039 --> 00:06:29,960
analysis type

189
00:06:27,639 --> 00:06:31,759
checking going a bit broader than Ruby

190
00:06:29,960 --> 00:06:33,880
as a language one big thing that Ruby

191
00:06:31,759 --> 00:06:35,280
has as a community that I wish python

192
00:06:33,880 --> 00:06:37,240
was better at is the Ruby for good

193
00:06:35,280 --> 00:06:39,520
organization it's a dedicated support

194
00:06:37,240 --> 00:06:41,280
structure for Ruby projects work with

195
00:06:39,520 --> 00:06:43,199
which work with specific nonprofit

196
00:06:41,280 --> 00:06:44,639
organizations we have a lot of paths for

197
00:06:43,199 --> 00:06:46,440
advancing python itself and some of the

198
00:06:44,639 --> 00:06:48,039
bigger languages like Jango but

199
00:06:46,440 --> 00:06:49,479
sometimes it's nice to do real tangible

200
00:06:48,039 --> 00:06:51,800
good in the

201
00:06:49,479 --> 00:06:53,240
world after Ruby the next logical place

202
00:06:51,800 --> 00:06:54,680
to look for ideas are the JavaScript and

203
00:06:53,240 --> 00:06:56,280
typescript worlds I'm not going to

204
00:06:54,680 --> 00:06:58,080
repeat every feature shared by all these

205
00:06:56,280 --> 00:07:00,240
languages like how JavaScript also has a

206
00:06:58,080 --> 00:07:01,560
great Anonymous function syntax we'd be

207
00:07:00,240 --> 00:07:02,960
here all day I'm just going to focus on

208
00:07:01,560 --> 00:07:04,919
some of the interesting

209
00:07:02,960 --> 00:07:06,599
bits this one has been discussed at

210
00:07:04,919 --> 00:07:08,000
Great length in the python Community but

211
00:07:06,599 --> 00:07:10,680
I personally think they should be added

212
00:07:08,000 --> 00:07:12,680
and it's my talk so one more time null

213
00:07:10,680 --> 00:07:14,479
cesing operators are very similar to an

214
00:07:12,680 --> 00:07:16,479
or but they're checking explicitly for

215
00:07:14,479 --> 00:07:19,280
nulles rather than any kind of falsy

216
00:07:16,479 --> 00:07:20,759
value safe navigation operators build on

217
00:07:19,280 --> 00:07:22,440
top of that you can access a deeply

218
00:07:20,759 --> 00:07:23,680
nested chain of fields but if anywhere

219
00:07:22,440 --> 00:07:25,440
there's anywhere in the chain there's a

220
00:07:23,680 --> 00:07:27,800
null the whole thing stops and just

221
00:07:25,440 --> 00:07:29,800
returns a null right there as you can

222
00:07:27,800 --> 00:07:31,199
see in this very small example safe

223
00:07:29,800 --> 00:07:32,720
navigation operators dramatically

224
00:07:31,199 --> 00:07:35,240
streamline code this is something I

225
00:07:32,720 --> 00:07:37,400
deeply miss when I'm working in

226
00:07:35,240 --> 00:07:39,440
Python typescript has always had a fine

227
00:07:37,400 --> 00:07:41,400
line to walk it applies type semantics

228
00:07:39,440 --> 00:07:43,080
to JavaScript but it can't fundamentally

229
00:07:41,400 --> 00:07:45,199
reshape the language or be too

230
00:07:43,080 --> 00:07:47,039
alienating so they knew the majority of

231
00:07:45,199 --> 00:07:48,280
object passed around within a codebase

232
00:07:47,039 --> 00:07:49,919
were going to be the JavaScript

233
00:07:48,280 --> 00:07:52,879
equivalent of just a plain dict no

234
00:07:49,919 --> 00:07:54,800
matter what they also wanted to allow

235
00:07:52,879 --> 00:07:56,680
typing information be gradually added to

236
00:07:54,800 --> 00:07:59,039
a code base out of this comes the

237
00:07:56,680 --> 00:07:59,960
synthesized requirement to support type

238
00:07:59,039 --> 00:08:02,560
DEC

239
00:07:59,960 --> 00:08:04,680
against arbitrary inline dicks in Python

240
00:08:02,560 --> 00:08:06,759
terms this is kind of similar to a data

241
00:08:04,680 --> 00:08:08,520
class but it's completely ad hoc and

242
00:08:06,759 --> 00:08:10,639
without requiring either the upstream or

243
00:08:08,520 --> 00:08:12,560
Downstream code to change this certainly

244
00:08:10,639 --> 00:08:14,319
isn't as nice as a fully typed code base

245
00:08:12,560 --> 00:08:16,319
but it gives a smoother on-ramp to doing

246
00:08:14,319 --> 00:08:18,720
one bit of code at a

247
00:08:16,319 --> 00:08:20,400
time I'm kind of cheating here we do

248
00:08:18,720 --> 00:08:22,400
have promises in python as part of the

249
00:08:20,400 --> 00:08:24,280
con Futures library and perhaps their

250
00:08:22,400 --> 00:08:26,639
moment is over anyway uh with the

251
00:08:24,280 --> 00:08:30,319
continued up uptake in explicit asyn A

252
00:08:26,639 --> 00:08:31,960
we as Chris so rightfully pointed out uh

253
00:08:30,319 --> 00:08:33,519
most of those apis are now what you'd

254
00:08:31,960 --> 00:08:35,519
use in Python and even in other

255
00:08:33,519 --> 00:08:36,719
languages too but they're still worth

256
00:08:35,519 --> 00:08:39,320
studying even as only part of our

257
00:08:36,719 --> 00:08:40,800
Collective history promise chaining is a

258
00:08:39,320 --> 00:08:42,599
compact way to handle multiple

259
00:08:40,800 --> 00:08:44,760
interacting concurrent operations

260
00:08:42,599 --> 00:08:46,120
without dedicated syntax I'm pulling my

261
00:08:44,760 --> 00:08:47,800
punches with this a little bit this is a

262
00:08:46,120 --> 00:08:49,279
really simplistic example and in

263
00:08:47,800 --> 00:08:51,080
practice many promise chains were a lot

264
00:08:49,279 --> 00:08:52,920
harder to read but perhaps the technique

265
00:08:51,080 --> 00:08:55,880
we'll find another application

266
00:08:52,920 --> 00:08:57,200
someday on the topic of concurrency go

267
00:08:55,880 --> 00:08:58,519
it's no secret that go structured

268
00:08:57,200 --> 00:09:00,360
concurrency has made it a frequent

269
00:08:58,519 --> 00:09:01,600
choice in complex system code and

270
00:09:00,360 --> 00:09:03,079
honestly I think Python's already

271
00:09:01,600 --> 00:09:05,240
heading in a lot of the same

272
00:09:03,079 --> 00:09:07,399
directions if you've seen anything about

273
00:09:05,240 --> 00:09:09,240
go it's probably this go routines are

274
00:09:07,399 --> 00:09:10,680
concurrency mechanism more or less the

275
00:09:09,240 --> 00:09:12,519
same as we have with async functions

276
00:09:10,680 --> 00:09:14,360
except they're truly concurrent and

277
00:09:12,519 --> 00:09:16,519
channels are cues to move small bits of

278
00:09:14,360 --> 00:09:18,360
data between them I think this model is

279
00:09:16,519 --> 00:09:20,079
about to get very interesting in python

280
00:09:18,360 --> 00:09:21,760
as we've seen the addition of both Gil

281
00:09:20,079 --> 00:09:23,600
isolated sub interpreters and the new no

282
00:09:21,760 --> 00:09:25,839
Gill mode there's a future where we get

283
00:09:23,600 --> 00:09:27,720
something very similar to the code on

284
00:09:25,839 --> 00:09:30,040
screen but with full multicore

285
00:09:27,720 --> 00:09:31,519
concurrency

286
00:09:30,040 --> 00:09:33,560
runtime type checks are something python

287
00:09:31,519 --> 00:09:35,200
has always kept at arms length would you

288
00:09:33,560 --> 00:09:36,839
technically have it there's that runtime

289
00:09:35,200 --> 00:09:39,720
checkable decorator that comes from the

290
00:09:36,839 --> 00:09:41,600
typing module uh but as the docs warn

291
00:09:39,720 --> 00:09:43,680
you very emphatically all this does is

292
00:09:41,600 --> 00:09:45,839
check that the right method names exist

293
00:09:43,680 --> 00:09:48,000
not their type signatures compare this

294
00:09:45,839 --> 00:09:49,399
To Go's runtime casts which do a more

295
00:09:48,000 --> 00:09:50,640
complete check and they can either fail

296
00:09:49,399 --> 00:09:52,760
with an error message or just straight

297
00:09:50,640 --> 00:09:54,800
up crash your program whichever you

298
00:09:52,760 --> 00:09:56,640
prefer this would requ that there would

299
00:09:54,800 --> 00:09:57,560
be something in cpython itself that

300
00:09:56,640 --> 00:09:59,680
understands how to compare type

301
00:09:57,560 --> 00:10:02,040
signatures but perhaps we've come enough

302
00:09:59,680 --> 00:10:02,920
in our type hinting journey to be ready

303
00:10:02,040 --> 00:10:05,720
for

304
00:10:02,920 --> 00:10:07,440
that in a totally different direction go

305
00:10:05,720 --> 00:10:09,920
is among the Best in Class alongside

306
00:10:07,440 --> 00:10:11,680
rust with preparing easy Standalone

307
00:10:09,920 --> 00:10:13,760
executables as well as crossplatform

308
00:10:11,680 --> 00:10:15,760
compilation in the server world this

309
00:10:13,760 --> 00:10:17,360
makes deployment really easy inside

310
00:10:15,760 --> 00:10:19,560
minimalist containers all you need from

311
00:10:17,360 --> 00:10:21,920
the OS is some TLS certificates maybe a

312
00:10:19,560 --> 00:10:23,560
time zone definition file or two and for

313
00:10:21,920 --> 00:10:25,480
command line tools you can distribute a

314
00:10:23,560 --> 00:10:27,560
single binary without concern for the

315
00:10:25,480 --> 00:10:29,360
system Library versions or anything else

316
00:10:27,560 --> 00:10:31,120
uh briefcase has relatively recently

317
00:10:29,360 --> 00:10:32,160
added some simple CLI packaging support

318
00:10:31,120 --> 00:10:35,279
and I'm hopeful that it will improve

319
00:10:32,160 --> 00:10:37,920
things here make Russ make this

320
00:10:35,279 --> 00:10:39,760
better uh unic kernels build on static

321
00:10:37,920 --> 00:10:41,040
compilation and take it a level deeper

322
00:10:39,760 --> 00:10:42,800
rather than linking in the language

323
00:10:41,040 --> 00:10:44,519
runtime and maybe libc and some other

324
00:10:42,800 --> 00:10:46,720
files what if we statically link an

325
00:10:44,519 --> 00:10:48,120
entire operating system unical

326
00:10:46,720 --> 00:10:49,760
Frameworks take your app code and they

327
00:10:48,120 --> 00:10:51,600
turn it into a bootable OS image

328
00:10:49,760 --> 00:10:52,360
designed to run in a hypervisor instead

329
00:10:51,600 --> 00:10:54,399
of

330
00:10:52,360 --> 00:10:56,279
Linux while containers kind of took the

331
00:10:54,399 --> 00:10:57,639
sales out of this technique uh it can

332
00:10:56,279 --> 00:10:58,800
still matter when you want to get every

333
00:10:57,639 --> 00:11:00,959
last bit of performance out of your

334
00:10:58,800 --> 00:11:03,000
Hardware while also ensuring absolute

335
00:11:00,959 --> 00:11:04,680
maximum security isolation there are

336
00:11:03,000 --> 00:11:05,920
tools out there like uncraft that can do

337
00:11:04,680 --> 00:11:07,519
this today but they're all general

338
00:11:05,920 --> 00:11:08,920
purpose tools they lack the Deep

339
00:11:07,519 --> 00:11:09,959
integration into python to really make

340
00:11:08,920 --> 00:11:11,880
them

341
00:11:09,959 --> 00:11:13,200
shine if you're only here for things

342
00:11:11,880 --> 00:11:14,800
that have even a tiny chance of

343
00:11:13,200 --> 00:11:15,680
happening in Python I wish you well but

344
00:11:14,800 --> 00:11:17,639
now I want to look at a little bit of

345
00:11:15,680 --> 00:11:19,360
the long taale weirder features you

346
00:11:17,639 --> 00:11:21,000
maybe have heard about but it's really

347
00:11:19,360 --> 00:11:22,360
unlikely that python will actually

348
00:11:21,000 --> 00:11:24,440
probably discuss adding these at any

349
00:11:22,360 --> 00:11:26,839
point uh even against those long odds I

350
00:11:24,440 --> 00:11:29,320
still think there are things here to

351
00:11:26,839 --> 00:11:32,320
learn on the complete opposite end from

352
00:11:29,320 --> 00:11:34,399
packaging is easy on liners Pearl has a

353
00:11:32,320 --> 00:11:36,320
lot of interesting Design Elements but

354
00:11:34,399 --> 00:11:38,519
by far the most useful for me day to-day

355
00:11:36,320 --> 00:11:40,480
is Pearl pie it's a sequence of command

356
00:11:38,519 --> 00:11:42,519
line arguments that says take this list

357
00:11:40,480 --> 00:11:44,560
of input files run this expression

358
00:11:42,519 --> 00:11:46,560
against every line of every file and

359
00:11:44,560 --> 00:11:48,880
then write the results back to those

360
00:11:46,560 --> 00:11:50,880
files taken together this allows for a

361
00:11:48,880 --> 00:11:53,320
super set of things like said and O and

362
00:11:50,880 --> 00:11:54,720
Bash but in an incredibly compact

363
00:11:53,320 --> 00:11:56,120
package and without the Myriad of

364
00:11:54,720 --> 00:11:58,600
Provider differences between all those

365
00:11:56,120 --> 00:11:59,880
tools that have occurred over time now

366
00:11:58,600 --> 00:12:01,880
making a python version of this in a

367
00:11:59,880 --> 00:12:03,560
library would be easy and they are out

368
00:12:01,880 --> 00:12:05,600
there you can use them but one of the

369
00:12:03,560 --> 00:12:07,279
real values of pearl pie you can drop me

370
00:12:05,600 --> 00:12:09,360
on to pretty much any server made from

371
00:12:07,279 --> 00:12:11,519
1995 onwards and there's probably going

372
00:12:09,360 --> 00:12:13,120
to be a pearl binary on there somewhere

373
00:12:11,519 --> 00:12:14,560
maybe if we added this to python in the

374
00:12:13,120 --> 00:12:17,279
standard library now it could be

375
00:12:14,560 --> 00:12:19,800
similarly Omni present in 10

376
00:12:17,279 --> 00:12:21,720
years link is short for language

377
00:12:19,800 --> 00:12:24,839
integrated query and it's basically what

378
00:12:21,720 --> 00:12:26,639
if SQL was our API this is most direct

379
00:12:24,839 --> 00:12:28,079
when it's used as a type of omm and that

380
00:12:26,639 --> 00:12:30,760
is definitely the most common use case

381
00:12:28,079 --> 00:12:32,839
for it but C took this even further and

382
00:12:30,760 --> 00:12:35,560
you can build arbitrary providers on top

383
00:12:32,839 --> 00:12:37,000
of link to make it do anything you want

384
00:12:35,560 --> 00:12:38,480
python has taken some baby steps down

385
00:12:37,000 --> 00:12:39,839
this path we have cross Library

386
00:12:38,480 --> 00:12:42,160
standards for data frames and you can

387
00:12:39,839 --> 00:12:44,320
put duck DB on top of those to get SQL

388
00:12:42,160 --> 00:12:45,920
like exploration uh but what if that was

389
00:12:44,320 --> 00:12:47,639
a language syntax that anything could

390
00:12:45,920 --> 00:12:50,079
plug

391
00:12:47,639 --> 00:12:51,519
into call with current continuations

392
00:12:50,079 --> 00:12:53,680
will take some explaining so just bear

393
00:12:51,519 --> 00:12:56,120
with me at a basic level it's a kind of

394
00:12:53,680 --> 00:12:57,720
non-local return like an exception and

395
00:12:56,120 --> 00:12:59,720
indeed we can do a very basic

396
00:12:57,720 --> 00:13:02,680
implementation on this in Python using

397
00:12:59,720 --> 00:13:04,320
exceptions so we take a function we want

398
00:13:02,680 --> 00:13:06,120
to run that function the function will

399
00:13:04,320 --> 00:13:08,279
receive one argument which is itself a

400
00:13:06,120 --> 00:13:10,440
callable when the function wishes to

401
00:13:08,279 --> 00:13:12,079
return it should call that callable with

402
00:13:10,440 --> 00:13:14,480
the parameter being that the value that

403
00:13:12,079 --> 00:13:16,320
it wants to return this alone would be

404
00:13:14,480 --> 00:13:17,959
pretty boring though the real power of

405
00:13:16,320 --> 00:13:20,120
continuations is that in some languages

406
00:13:17,959 --> 00:13:22,920
they can be moved around stored as

407
00:13:20,120 --> 00:13:24,279
variables or even serialized to disk all

408
00:13:22,920 --> 00:13:25,920
allowing you to jump right back to the

409
00:13:24,279 --> 00:13:27,399
point of the code where the continuation

410
00:13:25,920 --> 00:13:29,680
was generated this makes it something

411
00:13:27,399 --> 00:13:31,240
like a structured go-to imagine being

412
00:13:29,680 --> 00:13:32,839
able to save and restore the state of a

413
00:13:31,240 --> 00:13:34,279
generator function as if were any other

414
00:13:32,839 --> 00:13:35,320
object uh I wrote that line but you

415
00:13:34,279 --> 00:13:37,760
don't really have to imagine because

416
00:13:35,320 --> 00:13:39,560
Chris showed us how to do that but what

417
00:13:37,760 --> 00:13:41,560
if it wasn't like

418
00:13:39,560 --> 00:13:44,000
that

419
00:13:41,560 --> 00:13:46,000
um while continuations are mostly

420
00:13:44,000 --> 00:13:47,560
associated with scheme they are very

421
00:13:46,000 --> 00:13:49,800
effectively used in the small talk web

422
00:13:47,560 --> 00:13:51,600
framework Seaside in seide the whole

423
00:13:49,800 --> 00:13:54,480
state of the application is managed via

424
00:13:51,600 --> 00:13:56,399
pickable continuations so in that small

425
00:13:54,480 --> 00:13:58,680
talk code up there that invocation of

426
00:13:56,399 --> 00:14:00,519
Color Picker this is inside a web app so

427
00:13:58,680 --> 00:14:02,160
that with Color Picker is going to

428
00:14:00,519 --> 00:14:04,000
return an HTP response to the client

429
00:14:02,160 --> 00:14:06,040
telling some JavaScript to please open a

430
00:14:04,000 --> 00:14:08,560
modal showing a Color Picker wait for

431
00:14:06,040 --> 00:14:10,360
the user to do a thing get a response

432
00:14:08,560 --> 00:14:11,839
then send that back as another request

433
00:14:10,360 --> 00:14:13,880
up to the server where the server will

434
00:14:11,839 --> 00:14:16,360
unpickle the the continuation and pick

435
00:14:13,880 --> 00:14:18,040
up right where it left off in the code

436
00:14:16,360 --> 00:14:20,399
so this is all the benefits of the

437
00:14:18,040 --> 00:14:23,199
cleanliness of a single page client side

438
00:14:20,399 --> 00:14:25,560
app but in server side

439
00:14:23,199 --> 00:14:27,040
code and because I really love how small

440
00:14:25,560 --> 00:14:29,560
talk does flow control here's a really

441
00:14:27,040 --> 00:14:30,959
rough Port of some of it to python the

442
00:14:29,560 --> 00:14:32,759
short and important version of this

443
00:14:30,959 --> 00:14:36,079
small talk doesn't have special Syntax

444
00:14:32,759 --> 00:14:39,000
for flow control so for example if is a

445
00:14:36,079 --> 00:14:40,360
method of The Bu object and so is while

446
00:14:39,000 --> 00:14:43,000
there's another an example of each of

447
00:14:40,360 --> 00:14:44,800
those up there if you can read them uh

448
00:14:43,000 --> 00:14:46,199
in Practical implementations these don't

449
00:14:44,800 --> 00:14:47,759
actually happen they get compiled into

450
00:14:46,199 --> 00:14:49,600
special constructs because we want code

451
00:14:47,759 --> 00:14:51,480
to run quickly but it's a great example

452
00:14:49,600 --> 00:14:53,560
of doing more with less in language

453
00:14:51,480 --> 00:14:55,880
design and if you want to go even more

454
00:14:53,560 --> 00:14:57,160
minimal you can look at tiny fourths you

455
00:14:55,880 --> 00:14:58,560
can bootstrap a whole language

456
00:14:57,160 --> 00:15:01,079
interpreter with a few hundred bytes of

457
00:14:58,560 --> 00:15:02,560
assembly 20 to 30 keywords from those

458
00:15:01,079 --> 00:15:04,360
you build a basic parser and then an

459
00:15:02,560 --> 00:15:06,839
interpreter Loop and then some more

460
00:15:04,360 --> 00:15:08,000
control flow and then an entire program

461
00:15:06,839 --> 00:15:10,480
fourth Kernels have been written as

462
00:15:08,000 --> 00:15:12,240
small as 400 bytes of machine code

463
00:15:10,480 --> 00:15:14,000
minimalism isn't really a specific goal

464
00:15:12,240 --> 00:15:15,480
of python but it's a good reminder of

465
00:15:14,000 --> 00:15:17,240
how little is needed for a productive

466
00:15:15,480 --> 00:15:18,480
system what would a minimal python look

467
00:15:17,240 --> 00:15:20,480
like what's something that would be just

468
00:15:18,480 --> 00:15:21,720
powerful enough to write all of the rest

469
00:15:20,480 --> 00:15:23,880
of

470
00:15:21,720 --> 00:15:25,600
python another interesting blend of

471
00:15:23,880 --> 00:15:27,720
language features and Frameworks is LL

472
00:15:25,600 --> 00:15:29,920
Lang's open telephony platform or

473
00:15:27,720 --> 00:15:31,959
OTP speciic specifically the supervisor

474
00:15:29,920 --> 00:15:34,279
system earling programs are to use

475
00:15:31,959 --> 00:15:36,720
Python terms a series of interacting

476
00:15:34,279 --> 00:15:38,279
async tasks all running in parallel we

477
00:15:36,720 --> 00:15:40,440
can do that in Python today but what

478
00:15:38,279 --> 00:15:41,880
happens when one of those tasks crashes

479
00:15:40,440 --> 00:15:44,480
it's often difficult to sort out how to

480
00:15:41,880 --> 00:15:45,920
respond exactly OTP applies similar

481
00:15:44,480 --> 00:15:47,920
Concepts from the server and container

482
00:15:45,920 --> 00:15:49,639
management world but to individual tasks

483
00:15:47,920 --> 00:15:51,959
within a complex application we can set

484
00:15:49,639 --> 00:15:53,880
the supervisor to restart a failed task

485
00:15:51,959 --> 00:15:55,279
automatically or have a custom restart

486
00:15:53,880 --> 00:15:57,120
Behavior like waiting a certain amount

487
00:15:55,279 --> 00:15:59,720
of time or we can set up a dependency

488
00:15:57,120 --> 00:16:01,959
tree so that if any in the tree crashes

489
00:15:59,720 --> 00:16:04,040
it restarts the entire set to ensure

490
00:16:01,959 --> 00:16:05,759
consistency this is how llang builds

491
00:16:04,040 --> 00:16:07,519
incredibly High uptime systems like

492
00:16:05,759 --> 00:16:09,680
phone switches without magical

493
00:16:07,519 --> 00:16:11,680
programmers who don't write

494
00:16:09,680 --> 00:16:13,319
bugs this could mesh well with the sub

495
00:16:11,680 --> 00:16:15,040
interpreter work we looked at before and

496
00:16:13,319 --> 00:16:17,079
in fact Go's co- routines and channels

497
00:16:15,040 --> 00:16:18,759
were heavily influenced by otp's server

498
00:16:17,079 --> 00:16:20,120
and mailbox model there could be a

499
00:16:18,759 --> 00:16:21,680
future where we have something like task

500
00:16:20,120 --> 00:16:25,040
groups not just as a synchronization

501
00:16:21,680 --> 00:16:27,120
mechanism but also for runtime

502
00:16:25,040 --> 00:16:28,560
supervision python has made incredible

503
00:16:27,120 --> 00:16:29,839
inroads as being the language of choice

504
00:16:28,560 --> 00:16:31,639
for s scientific Computing and data

505
00:16:29,839 --> 00:16:33,360
science but in almost every one of those

506
00:16:31,639 --> 00:16:35,319
applications you'll find Fortran under

507
00:16:33,360 --> 00:16:37,120
the hood most people think of Fortran as

508
00:16:35,319 --> 00:16:39,639
part of our technical past but it's very

509
00:16:37,120 --> 00:16:41,680
much alive and for good reason forr is

510
00:16:39,639 --> 00:16:43,639
almost perfectly designed to be able to

511
00:16:41,680 --> 00:16:45,519
write complex mathematical algorithms

512
00:16:43,639 --> 00:16:47,680
which can be compiled to the fastest

513
00:16:45,519 --> 00:16:49,319
possible implementation on each platform

514
00:16:47,680 --> 00:16:51,040
so rather than maintaining a separate

515
00:16:49,319 --> 00:16:53,959
linear algebra implementation in

516
00:16:51,040 --> 00:16:55,880
assembly on every OS we have a single

517
00:16:53,959 --> 00:16:58,319
one in Fortran and we trust in the tool

518
00:16:55,880 --> 00:17:00,319
chain even C can't compile things as

519
00:16:58,319 --> 00:17:02,360
effectively 4 Trend in a lot of cases

520
00:17:00,319 --> 00:17:04,799
despite most of us thinking of C as the

521
00:17:02,360 --> 00:17:06,839
lowest level thing out there specialized

522
00:17:04,799 --> 00:17:09,039
use cases need specialized

523
00:17:06,839 --> 00:17:11,160
tools and while Fortran is still alive

524
00:17:09,039 --> 00:17:12,720
and well cobal is mostly though not

525
00:17:11,160 --> 00:17:14,799
entirely in the rearview mirror at this

526
00:17:12,720 --> 00:17:16,120
point but like we talked about with PHP

527
00:17:14,799 --> 00:17:17,760
it's fun to look at language features

528
00:17:16,120 --> 00:17:19,919
that grow from a tool with a single

529
00:17:17,760 --> 00:17:21,360
primary purpose in Cobalt I think the

530
00:17:19,919 --> 00:17:24,199
most interesting part is the data

531
00:17:21,360 --> 00:17:26,439
division in modern terms every Cobalt

532
00:17:24,199 --> 00:17:28,199
program carries around the schemas of

533
00:17:26,439 --> 00:17:29,520
the databases it expects to read from

534
00:17:28,199 --> 00:17:30,640
and write to

535
00:17:29,520 --> 00:17:32,280
while these days it would more often be

536
00:17:30,640 --> 00:17:33,880
a network API rather than a structured

537
00:17:32,280 --> 00:17:35,400
file on disk We have basically the same

538
00:17:33,880 --> 00:17:37,559
problems to solve with producers and

539
00:17:35,400 --> 00:17:38,799
consumers of data agreeing on schemas

540
00:17:37,559 --> 00:17:40,840
while still allowing that schema to

541
00:17:38,799 --> 00:17:42,520
evolve over time Cobalt's implementation

542
00:17:40,840 --> 00:17:44,200
isn't perfect it comes from a time where

543
00:17:42,520 --> 00:17:46,120
those schema changes were expected to

544
00:17:44,200 --> 00:17:47,440
take months instead of hours but I've

545
00:17:46,120 --> 00:17:49,160
worked on a lot of projects that are

546
00:17:47,440 --> 00:17:50,799
decades newer and have substantially

547
00:17:49,160 --> 00:17:53,000
worse schem of

548
00:17:50,799 --> 00:17:54,559
management homomorphic encryption is

549
00:17:53,000 --> 00:17:56,120
probably the most out there of all these

550
00:17:54,559 --> 00:17:58,600
it's barely even a programming feature

551
00:17:56,120 --> 00:18:00,919
and more of a cool math trick the idea

552
00:17:58,600 --> 00:18:02,400
is doing computation on encrypted values

553
00:18:00,919 --> 00:18:04,200
without decrypting them so this means

554
00:18:02,400 --> 00:18:06,000
that you have data processing code that

555
00:18:04,200 --> 00:18:07,880
takes encrypted data that it does not

556
00:18:06,000 --> 00:18:10,000
have the decryption keys

557
00:18:07,880 --> 00:18:11,440
for the core way that this works is

558
00:18:10,000 --> 00:18:13,919
instead of thinking of your code AS

559
00:18:11,440 --> 00:18:15,520
lines of code compiling to machine

560
00:18:13,919 --> 00:18:17,159
instructions that are evaluated on some

561
00:18:15,520 --> 00:18:18,760
kind of virtual machine we're instead

562
00:18:17,159 --> 00:18:21,080
going to compile your code into

563
00:18:18,760 --> 00:18:22,600
effectively a virtual Asic digital

564
00:18:21,080 --> 00:18:24,960
circuit Gates that'll Implement a

565
00:18:22,600 --> 00:18:26,240
homeomorphic encryption flow given those

566
00:18:24,960 --> 00:18:27,840
constraints there are of course very

567
00:18:26,240 --> 00:18:29,280
hard limitations in the kind of

568
00:18:27,840 --> 00:18:31,799
computation that you can do and also

569
00:18:29,280 --> 00:18:33,320
it's still pretty slow but in an

570
00:18:31,799 --> 00:18:34,640
increasingly privacy conscious World

571
00:18:33,320 --> 00:18:35,880
maybe this is worth more of a look from

572
00:18:34,640 --> 00:18:38,480
all of

573
00:18:35,880 --> 00:18:40,000
us I started off by saying I wasn't

574
00:18:38,480 --> 00:18:42,080
aiming for you to go off and turn these

575
00:18:40,000 --> 00:18:44,400
ideas into peps and while that is true I

576
00:18:42,080 --> 00:18:45,480
wish it wasn't a few of these ideas

577
00:18:44,400 --> 00:18:47,400
could probably be prototyped in

578
00:18:45,480 --> 00:18:49,320
libraries and gain some traction there

579
00:18:47,400 --> 00:18:51,480
but most of these to really succeed they

580
00:18:49,320 --> 00:18:52,679
would need a core language change the

581
00:18:51,480 --> 00:18:54,200
process for that has improved over the

582
00:18:52,679 --> 00:18:56,360
years but the steering Council has to

583
00:18:54,200 --> 00:18:58,440
walk a really fine line adding anything

584
00:18:56,360 --> 00:19:00,720
into cpython is a big commitment it

585
00:18:58,440 --> 00:19:02,559
can't just be experimental and it'll go

586
00:19:00,720 --> 00:19:04,440
away in a little bit so this leaves us

587
00:19:02,559 --> 00:19:05,799
in kind of a bind to get peer support

588
00:19:04,440 --> 00:19:07,640
for a change we need to prove that it's

589
00:19:05,799 --> 00:19:08,919
a good idea and for that we need people

590
00:19:07,640 --> 00:19:10,240
to be able to play with it but for

591
00:19:08,919 --> 00:19:12,039
people to be able to play with it we

592
00:19:10,240 --> 00:19:13,400
need to get a Core change in and we

593
00:19:12,039 --> 00:19:14,480
can't do that until we have peer support

594
00:19:13,400 --> 00:19:17,120
and now we're

595
00:19:14,480 --> 00:19:18,600
stuck macros are a language construct to

596
00:19:17,120 --> 00:19:20,240
allow metaprogramming inside the

597
00:19:18,600 --> 00:19:21,840
language itself adding a new

598
00:19:20,240 --> 00:19:23,480
experimental syntax or changing some

599
00:19:21,840 --> 00:19:26,039
syntax that's already there or writing

600
00:19:23,480 --> 00:19:27,400
your code in rout 13 the sky the limit

601
00:19:26,039 --> 00:19:30,159
the two best places to look for ideas

602
00:19:27,400 --> 00:19:31,600
about macros are rust and lisp C does

603
00:19:30,159 --> 00:19:33,120
heavily feature macros but it takes a

604
00:19:31,600 --> 00:19:34,799
much less structured approach treating

605
00:19:33,120 --> 00:19:37,000
source code more as just text to be

606
00:19:34,799 --> 00:19:38,919
templated uh and that does technically

607
00:19:37,000 --> 00:19:40,960
work but Lis and rust show us a deeper

608
00:19:38,919 --> 00:19:43,200
power defining macros in terms of the

609
00:19:40,960 --> 00:19:44,720
symbols of a language but not its syntax

610
00:19:43,200 --> 00:19:47,080
so that we can create our

611
00:19:44,720 --> 00:19:50,360
own python does have a proposed macro

612
00:19:47,080 --> 00:19:52,039
system heavily influenced by rusts but

613
00:19:50,360 --> 00:19:53,600
it has fallen into the same trap as we

614
00:19:52,039 --> 00:19:54,960
were trying to escape by having a macro

615
00:19:53,600 --> 00:19:56,720
system in the first place and has

616
00:19:54,960 --> 00:19:59,360
received little discussion over the

617
00:19:56,720 --> 00:20:00,880
years I want to talk about lisp does

618
00:19:59,360 --> 00:20:02,480
macros before we can do that we need to

619
00:20:00,880 --> 00:20:04,039
talk about how lisp code is structured

620
00:20:02,480 --> 00:20:05,720
and before that we need to talk about

621
00:20:04,039 --> 00:20:07,799
how python code is structured so like

622
00:20:05,720 --> 00:20:09,679
most other languages Python's parer

623
00:20:07,799 --> 00:20:11,360
produces an abstract syntax tree this

624
00:20:09,679 --> 00:20:13,360
isn't bite code this is a highly

625
00:20:11,360 --> 00:20:15,799
structured representation of your code

626
00:20:13,360 --> 00:20:17,559
text and it's ready to be handed off to

627
00:20:15,799 --> 00:20:20,000
the compiler and other layers for

628
00:20:17,559 --> 00:20:21,960
forther execution or or

629
00:20:20,000 --> 00:20:24,400
optimization lisp does have something

630
00:20:21,960 --> 00:20:26,600
like that but it's way simpler lisp has

631
00:20:24,400 --> 00:20:28,840
almost no actual syntax everything in

632
00:20:26,600 --> 00:20:30,080
lisp is one of two things an atom which

633
00:20:28,840 --> 00:20:32,720
is their word for a symbol which we

634
00:20:30,080 --> 00:20:34,919
talked about before or a pair of two

635
00:20:32,720 --> 00:20:36,679
things Each of which can again be either

636
00:20:34,919 --> 00:20:39,000
an atom or a pair of two

637
00:20:36,679 --> 00:20:40,799
things this is just the barest bit of

638
00:20:39,000 --> 00:20:43,080
structure but it also means that macros

639
00:20:40,799 --> 00:20:44,400
and lisp can do almost anything as long

640
00:20:43,080 --> 00:20:46,320
as the input and output are both s

641
00:20:44,400 --> 00:20:47,799
expressions and basically everything can

642
00:20:46,320 --> 00:20:49,960
be an S

643
00:20:47,799 --> 00:20:52,559
expression the way def macro Works in

644
00:20:49,960 --> 00:20:54,440
lisp so we've parsed the code into an S

645
00:20:52,559 --> 00:20:56,159
expression and something is going to

646
00:20:54,440 --> 00:20:58,640
start spidering out over that code

647
00:20:56,159 --> 00:21:01,320
looking for special atoms def macro it's

648
00:20:58,640 --> 00:21:02,840
self is a special atom and it takes the

649
00:21:01,320 --> 00:21:05,520
name the parameters and the body of the

650
00:21:02,840 --> 00:21:07,640
macro and it registers that name as a

651
00:21:05,520 --> 00:21:09,480
special atom and it's going to continue

652
00:21:07,640 --> 00:21:11,440
spidering out of the code if it sees

653
00:21:09,480 --> 00:21:13,679
that name it'll say it's a special atom

654
00:21:11,440 --> 00:21:15,640
I know what to do I will go find the

655
00:21:13,679 --> 00:21:17,000
body of that macro I will take the

656
00:21:15,640 --> 00:21:18,960
arguments which remember are still

657
00:21:17,000 --> 00:21:21,480
symbolic the code isn't running it's

658
00:21:18,960 --> 00:21:23,400
just been parsed so we have the symbolic

659
00:21:21,480 --> 00:21:25,279
arguments we'll slam them into something

660
00:21:23,400 --> 00:21:27,279
we'll evaluate the body of the macro

661
00:21:25,279 --> 00:21:28,720
we'll get back another expression and

662
00:21:27,279 --> 00:21:31,360
we'll just plug it right in where

663
00:21:28,720 --> 00:21:33,480
macro's special atom

664
00:21:31,360 --> 00:21:35,440
was the important takeaway here is that

665
00:21:33,480 --> 00:21:37,039
macros break the usual flow of things

666
00:21:35,440 --> 00:21:39,120
usually our code is going to move

667
00:21:37,039 --> 00:21:41,039
linearly from tokenizing to parsing to

668
00:21:39,120 --> 00:21:42,520
compiling to executing there's a couple

669
00:21:41,039 --> 00:21:44,400
of newer steps in there you'll see jit

670
00:21:42,520 --> 00:21:46,120
compilers and stuff but the key with

671
00:21:44,400 --> 00:21:48,880
macros is they put a little subl Loop

672
00:21:46,120 --> 00:21:50,320
inside the parser step compiling and

673
00:21:48,880 --> 00:21:52,200
executing the body of the macro and

674
00:21:50,320 --> 00:21:53,960
returning the output to the parser this

675
00:21:52,200 --> 00:21:55,400
may seem like a minor thing but it's the

676
00:21:53,960 --> 00:21:57,320
pressure relief valve that lets us

677
00:21:55,400 --> 00:22:00,480
extend the parser without explicit

678
00:21:57,320 --> 00:22:02,320
support from the language

679
00:22:00,480 --> 00:22:04,080
rust has two main types of macros

680
00:22:02,320 --> 00:22:05,559
procedural macros are the equivalent of

681
00:22:04,080 --> 00:22:07,159
what we saw in lisp uh there's a couple

682
00:22:05,559 --> 00:22:08,880
flavors depending on exactly what kind

683
00:22:07,159 --> 00:22:10,919
of parsing you want to hook into but

684
00:22:08,880 --> 00:22:12,559
they all take a stream of tokens and

685
00:22:10,919 --> 00:22:14,400
they return a stream of tokens which

686
00:22:12,559 --> 00:22:15,559
means they can do basically anything

687
00:22:14,400 --> 00:22:17,360
they can even operate on things that

688
00:22:15,559 --> 00:22:19,159
aren't valid rust syntax in the first

689
00:22:17,360 --> 00:22:20,480
place and that's cool sometimes but most

690
00:22:19,159 --> 00:22:22,559
of the time when we're writing a rust

691
00:22:20,480 --> 00:22:24,559
macro we just want to slightly tweak the

692
00:22:22,559 --> 00:22:25,919
syntax that's already there and we use

693
00:22:24,559 --> 00:22:27,120
declarative macros with this thing

694
00:22:25,919 --> 00:22:30,080
called macro

695
00:22:27,120 --> 00:22:31,960
rules declar macros are very similar to

696
00:22:30,080 --> 00:22:33,760
Python's match case statements but again

697
00:22:31,960 --> 00:22:35,840
the code isn't running it's just been

698
00:22:33,760 --> 00:22:38,400
parsed so the parameters are symbolic

699
00:22:35,840 --> 00:22:41,000
expressions or types and the code block

700
00:22:38,400 --> 00:22:42,760
is returned rather than being executed

701
00:22:41,000 --> 00:22:44,760
it does support a little bit of like

702
00:22:42,760 --> 00:22:46,400
templating expansion basically although

703
00:22:44,760 --> 00:22:48,440
I can't really go into the full power of

704
00:22:46,400 --> 00:22:50,799
this in 30 seconds so hopefully this

705
00:22:48,440 --> 00:22:52,320
gives you just the barest idea the core

706
00:22:50,799 --> 00:22:54,039
idea is that basically anything can be

707
00:22:52,320 --> 00:22:56,880
prototyped inside a macro without

708
00:22:54,039 --> 00:23:00,200
explicit support from the rust

709
00:22:56,880 --> 00:23:01,520
team so why at all of this personally I

710
00:23:00,200 --> 00:23:03,279
just think it's fun to see what else is

711
00:23:01,520 --> 00:23:04,720
out there but if that's not enough I

712
00:23:03,279 --> 00:23:06,640
certainly hope that learning about these

713
00:23:04,720 --> 00:23:09,120
patterns and systems can help you better

714
00:23:06,640 --> 00:23:10,640
organize your code and apis in python or

715
00:23:09,120 --> 00:23:11,919
at least shows you what other tools are

716
00:23:10,640 --> 00:23:13,960
out there and in the end that makes us

717
00:23:11,919 --> 00:23:16,559
all better developers thank you very

718
00:23:13,960 --> 00:23:16,559
much any

719
00:23:19,080 --> 00:23:24,200
questions thank you Noah we do have time

720
00:23:22,000 --> 00:23:24,200
for

721
00:23:26,200 --> 00:23:30,440
questions okay we have a we have a

722
00:23:28,159 --> 00:23:30,440
question

723
00:23:30,960 --> 00:23:35,919
hello um of those lists you mentioned

724
00:23:34,400 --> 00:23:38,159
this is python ideas but not python

725
00:23:35,919 --> 00:23:40,279
ideas which of those python ideas do you

726
00:23:38,159 --> 00:23:42,440
think would be most likely to be

727
00:23:40,279 --> 00:23:43,880
implemented or to to to actually land in

728
00:23:42,440 --> 00:23:46,440
core and which one do you think would be

729
00:23:43,880 --> 00:23:48,360
the highest yield if it were reflecting

730
00:23:46,440 --> 00:23:49,840
on Chris's comments about the fact that

731
00:23:48,360 --> 00:23:52,360
generators were implemented which then

732
00:23:49,840 --> 00:23:54,960
made all these other things possible

733
00:23:52,360 --> 00:23:57,120
um I think safe navigation I mean I

734
00:23:54,960 --> 00:23:58,559
wasn't kidding that I think they are it

735
00:23:57,120 --> 00:24:00,480
is it is a mistake that we have not

736
00:23:58,559 --> 00:24:03,840
included them um I think that would

737
00:24:00,480 --> 00:24:05,320
probably be the highest value one uh it

738
00:24:03,840 --> 00:24:08,039
it makes it much more likely that people

739
00:24:05,320 --> 00:24:11,520
will write good error checking in deeply

740
00:24:08,039 --> 00:24:13,080
Nest exploring deeply nested objects um

741
00:24:11,520 --> 00:24:14,320
which is a thing that a lot of code

742
00:24:13,080 --> 00:24:16,480
these days does especially as we're

743
00:24:14,320 --> 00:24:18,240
trying to push more people into using

744
00:24:16,480 --> 00:24:20,840
structured typing things like data

745
00:24:18,240 --> 00:24:22,760
classes or adders we can't just call do

746
00:24:20,840 --> 00:24:25,200
getet on there and assume that it'll

747
00:24:22,760 --> 00:24:26,360
figure itself out um so I think safe

748
00:24:25,200 --> 00:24:29,440
naav operators for the second one the

749
00:24:26,360 --> 00:24:33,880
one that is most likely I I don't even

750
00:24:29,440 --> 00:24:37,679
know anymore uh the the the core team is

751
00:24:33,880 --> 00:24:37,679
an inscrutable box to a lot of

752
00:24:42,080 --> 00:24:47,720
us uh somebody is pointing that way

753
00:24:44,840 --> 00:24:47,720
pointing

754
00:24:55,080 --> 00:24:59,679
yep yeah hi um so you talked about what

755
00:24:58,399 --> 00:25:01,240
you'd like to see added to the language

756
00:24:59,679 --> 00:25:02,919
I'm interested if you've got any ideas

757
00:25:01,240 --> 00:25:04,720
of things you'd like to see removed

758
00:25:02,919 --> 00:25:07,039
maybe something that's non-standard

759
00:25:04,720 --> 00:25:08,600
compared to other languages that you'd

760
00:25:07,039 --> 00:25:10,679
like to see standardized across the

761
00:25:08,600 --> 00:25:13,080
board

762
00:25:10,679 --> 00:25:14,480
uh in in Practical terms I would like

763
00:25:13,080 --> 00:25:18,279
nothing removed because I remember the

764
00:25:14,480 --> 00:25:20,480
Python 3 uh migration period and

765
00:25:18,279 --> 00:25:22,880
breaking compat like that should never

766
00:25:20,480 --> 00:25:24,279
happen again um and I'm pretty sure that

767
00:25:22,880 --> 00:25:26,799
we have all convinced ourselves that it

768
00:25:24,279 --> 00:25:28,640
never will again so in Practical terms I

769
00:25:26,799 --> 00:25:31,840
think anything that has been out there

770
00:25:28,640 --> 00:25:34,760
that ship has sailed

771
00:25:31,840 --> 00:25:37,480
um if I if I could wave a magic wand

772
00:25:34,760 --> 00:25:38,600
though uh the one I actually ask that

773
00:25:37,480 --> 00:25:40,760
question when I'm doing interviews the

774
00:25:38,600 --> 00:25:42,880
big one would be packaging uh I think no

775
00:25:40,760 --> 00:25:45,200
one is under any impression that python

776
00:25:42,880 --> 00:25:49,240
packaging is in an amazing State um I

777
00:25:45,200 --> 00:25:52,039
would make that not I would I would make

778
00:25:49,240 --> 00:25:54,399
python push all of that into a better

779
00:25:52,039 --> 00:25:55,720
Direction um if I could remove one thing

780
00:25:54,399 --> 00:25:57,760
it would probably be multiple

781
00:25:55,720 --> 00:26:00,480
inheritance I think it does more harm

782
00:25:57,760 --> 00:26:02,919
than good and always has um class

783
00:26:00,480 --> 00:26:04,960
decorators should have been the correct

784
00:26:02,919 --> 00:26:06,880
solution there and I think most people

785
00:26:04,960 --> 00:26:08,480
these days they use that instead but

786
00:26:06,880 --> 00:26:11,240
multiple inheritance just breaks a lot

787
00:26:08,480 --> 00:26:13,279
of stuff in mysterious ways uh my

788
00:26:11,240 --> 00:26:15,080
understanding of that is that g read a

789
00:26:13,279 --> 00:26:16,559
paper thought it was really cool and

790
00:26:15,080 --> 00:26:18,600
implemented it that night and it was

791
00:26:16,559 --> 00:26:20,440
early on enough in the the python

792
00:26:18,600 --> 00:26:23,080
ecosystem that no one could tell him no

793
00:26:20,440 --> 00:26:25,080
so he just merged it um basically

794
00:26:23,080 --> 00:26:26,600
because he thought it was shiny which

795
00:26:25,080 --> 00:26:29,600
not really how we do language design

796
00:26:26,600 --> 00:26:29,600
anymore

797
00:26:30,159 --> 00:26:35,399
hi I'm one of the duck DB devs I'm

798
00:26:32,200 --> 00:26:37,120
curious to see what you would what you

799
00:26:35,399 --> 00:26:39,039
think that would look like integrating

800
00:26:37,120 --> 00:26:41,600
duck DB or another SQL engine into

801
00:26:39,039 --> 00:26:45,159
python itself uh probably a lot like it

802
00:26:41,600 --> 00:26:46,679
does in C um you know link there is a

803
00:26:45,159 --> 00:26:48,840
special keyword I think it is literally

804
00:26:46,679 --> 00:26:50,360
just select because they wanted to model

805
00:26:48,840 --> 00:26:53,039
it off SQL because that's what most

806
00:26:50,360 --> 00:26:54,880
people know and once it sees that

807
00:26:53,039 --> 00:26:57,000
keyword it moves into an alternate

808
00:26:54,880 --> 00:27:00,559
parsing mode that again very closely

809
00:26:57,000 --> 00:27:02,320
resembles SQL and it would just probably

810
00:27:00,559 --> 00:27:04,840
the simplest version is just that would

811
00:27:02,320 --> 00:27:06,840
be a very fancy quoting syntax that

812
00:27:04,840 --> 00:27:08,000
would just hand that off to something

813
00:27:06,840 --> 00:27:09,760
else there would be a provider

814
00:27:08,000 --> 00:27:12,279
registered um we've talked about this I

815
00:27:09,760 --> 00:27:14,640
think uh a bit with tagged strings of

816
00:27:12,279 --> 00:27:17,120
various types um being able to register

817
00:27:14,640 --> 00:27:20,080
like we have F strings and R strings you

818
00:27:17,120 --> 00:27:23,440
could have a duck string um and that

819
00:27:20,080 --> 00:27:25,080
would certainly be a lot closer uh but I

820
00:27:23,440 --> 00:27:27,440
I like things like typechecking and it

821
00:27:25,080 --> 00:27:29,600
would be a lot closer to being a core

822
00:27:27,440 --> 00:27:32,640
language Fe if it was more deeply

823
00:27:29,600 --> 00:27:32,640
exposed as part of the

824
00:27:37,360 --> 00:27:41,159
language thank you very much I think

825
00:27:39,399 --> 00:27:45,640
that's

826
00:27:41,159 --> 00:27:45,640
it give it up to Noah