WEBVTT 1 00:00:00.160 --> 00:00:03.720 So if I told you to design a tool for 2 00:00:03.759 --> 00:00:08.759 a sailor in the seventeen hundreds to prevent a deadly disease, 3 00:00:08.960 --> 00:00:09.759 you'd probably give. 4 00:00:09.640 --> 00:00:12.039 Them a lemon, right, a literal lemon, exactly. 5 00:00:12.400 --> 00:00:14.480 But if I told you to design a tool to 6 00:00:14.599 --> 00:00:18.039 prevent a modern software engineer from completely losing their mind, 7 00:00:18.440 --> 00:00:20.640 you'd give them, well, Haskell. 8 00:00:20.760 --> 00:00:23.199 Which sounds completely absurd on its face. 9 00:00:23.320 --> 00:00:26.199 I know, it really does. I mean, when you hear 10 00:00:26.280 --> 00:00:30.199 the phrase functional programming or you know Haskell, there is 11 00:00:30.280 --> 00:00:33.000 this immediate reflex to just tense up. 12 00:00:33.079 --> 00:00:33.719 Oh totally. 13 00:00:33.759 --> 00:00:36.520 It has this towering reputation, Yeah, like it's the mount 14 00:00:36.560 --> 00:00:40.799 Everest of coding, this hyper academic, mystical language built by 15 00:00:40.920 --> 00:00:42.320 and for math geniuses. 16 00:00:42.320 --> 00:00:44.759 Well what if that reputation is completely backwards? 17 00:00:44.880 --> 00:00:47.719 Right? What is a language supposedly built for these towering 18 00:00:47.759 --> 00:00:52.079 intellects is actually designed like perfectly for our flawed, limited 19 00:00:52.159 --> 00:00:52.840 human brains. 20 00:00:52.920 --> 00:00:54.520 That's exactly what we're looking at today. 21 00:00:54.840 --> 00:00:57.920 Welcome to today's deep dive. Our mission here is to unpack 22 00:00:57.960 --> 00:01:01.840 this notoriously intimidating subject and show you exactly how it's 23 00:01:01.880 --> 00:01:04.000 built to save you from information overload. 24 00:01:04.319 --> 00:01:07.799 And we are pulling from a truly unique source to 25 00:01:07.920 --> 00:01:11.480 do that. We're diving into excerpts from the book Haskell 26 00:01:11.599 --> 00:01:16.159 Programming from First Principles by Christopher Allen and Julie Marnuki. 27 00:01:15.959 --> 00:01:19.239 Which I have to say has gotten some extraordinary praise. 28 00:01:19.040 --> 00:01:22.519 Oh, unbelievable praise. You have reader feedback from people like 29 00:01:22.680 --> 00:01:25.200 Jason Kurt saying you know, there will forever be a 30 00:01:25.239 --> 00:01:27.159 Haskell before and Haskell after. 31 00:01:27.519 --> 00:01:28.519 That's quite the endorsement. 32 00:01:28.719 --> 00:01:31.400 Yeah, And you even have physicist David Deutsch praising it. 33 00:01:31.519 --> 00:01:33.719 He noted that it acts like a really good teacher 34 00:01:33.760 --> 00:01:36.280 who never presume is what you already. 35 00:01:35.920 --> 00:01:39.319 Know, right, so the right information just surfaces precisely when 36 00:01:39.319 --> 00:01:42.200 you need it exactly. And that is the exact tone 37 00:01:42.239 --> 00:01:44.560 we are bringing to you today. You do not need 38 00:01:44.599 --> 00:01:48.040 a computer science degree or a math background to follow along. 39 00:01:47.760 --> 00:01:49.040 With this, not at all. 40 00:01:49.120 --> 00:01:52.920 We are just exploring how complex ideas can be broken 41 00:01:53.000 --> 00:01:57.040 down to protect your cognitive load. And honestly, the story 42 00:01:57.040 --> 00:02:00.359 of how this book even came into existence sets that 43 00:02:00.480 --> 00:02:01.000 up perfectly. 44 00:02:01.239 --> 00:02:03.719 It really does, because it wasn't written by you know, 45 00:02:03.840 --> 00:02:06.640 two Iver Tower professors coding in a basement since the 46 00:02:06.760 --> 00:02:07.599 nineteen seventies. 47 00:02:07.719 --> 00:02:09.759 Right. The authorship dynamic is pretty fascinating. 48 00:02:09.919 --> 00:02:13.400 It explains the accessibility of the entire text. Really so. 49 00:02:13.599 --> 00:02:17.080 Chris Allen was a veteran programmer of like fifteen years. 50 00:02:17.159 --> 00:02:18.039 He knew what he was doing. 51 00:02:18.120 --> 00:02:21.240 Oh yeah, he had worked extensively in languages like Closure 52 00:02:21.319 --> 00:02:24.560 and Python, but he found himself hitting a wall trying 53 00:02:24.560 --> 00:02:26.919 to teach Haskell to beginners. 54 00:02:26.439 --> 00:02:28.439 Which is where Julie comes in exactly. 55 00:02:28.879 --> 00:02:32.199 His co author, Julie Moronuki was a linguist and when 56 00:02:32.240 --> 00:02:36.800 they began this project, she had absolutely zero programming experience wait, 57 00:02:37.039 --> 00:02:39.719 literally none, literally none. She had to learn what a 58 00:02:39.759 --> 00:02:42.919 compiler was, what a side effect was, what a stack 59 00:02:43.039 --> 00:02:46.000 overflow meant, entirely from scratch. 60 00:02:46.240 --> 00:02:49.280 That is wild, but I guess they teamed up because 61 00:02:49.560 --> 00:02:54.599 existing learning materials for Haskell relied almost entirely on analogies 62 00:02:54.639 --> 00:02:56.000 to other programming. 63 00:02:55.639 --> 00:02:58.599 Languages, right, which is a huge barrier. If a tutorial says, oh, 64 00:02:58.639 --> 00:03:00.719 this is just like a Python dictionary and you don't know. 65 00:03:00.719 --> 00:03:02.319 Python, you're just locked out of the garden. 66 00:03:02.479 --> 00:03:06.680 Exactly. So Julie became the ultimate stress tester for the book. 67 00:03:06.800 --> 00:03:08.039 That makes so much sense. 68 00:03:08.159 --> 00:03:11.719 If a concept relied on a faulty analogy or assumed 69 00:03:11.759 --> 00:03:15.360 prior programming knowledge, she flagged it, and they just rebuilt 70 00:03:15.360 --> 00:03:16.840 the explanation from the ground up. 71 00:03:17.039 --> 00:03:19.080 Okay, but I want to play Devil's advocate here for 72 00:03:19.080 --> 00:03:19.520 a second. 73 00:03:19.560 --> 00:03:20.000 Go for it. 74 00:03:20.159 --> 00:03:22.960 If this language is so brilliant and this book proves 75 00:03:22.960 --> 00:03:25.960 it can be made accessible, why does Haskell have this 76 00:03:26.120 --> 00:03:30.800 wild internet rumor that you need a PhD in I 77 00:03:30.800 --> 00:03:34.919 don't know, monodology, Yeah, just to write a simple Hello 78 00:03:35.000 --> 00:03:35.919 World program. 79 00:03:36.120 --> 00:03:37.319 Yeah, that's the big question. 80 00:03:37.479 --> 00:03:39.960 Is it just academic elitism keeping people away? 81 00:03:40.319 --> 00:03:43.120 To understand why the tech industry sometimes shies away from 82 00:03:43.120 --> 00:03:47.159 its best tools, the authors use a really fascinating historical lens. 83 00:03:47.520 --> 00:03:50.400 They actually compare the adoption of programming languages to the 84 00:03:50.439 --> 00:03:51.520 history of scurvy. 85 00:03:51.680 --> 00:03:55.080 Okay, scurvy, the disease that wiped out pirates and sailors. 86 00:03:55.199 --> 00:03:58.159 That's the one between the year's fifteen hundred and eighteen hundred. 87 00:03:58.319 --> 00:04:01.159 Scurvy killed and estimated two two million sailors. 88 00:04:01.319 --> 00:04:03.039 Oh wow, two millions. 89 00:04:03.080 --> 00:04:05.840 Yeah, it was devastating. But the cure citrus was actually 90 00:04:05.879 --> 00:04:08.639 recommended by the Surgeon General of the East India Company 91 00:04:08.680 --> 00:04:10.400 as early as sixteen fourteen. 92 00:04:10.479 --> 00:04:12.599 It's sixteen fourteen, so they know about it way early. 93 00:04:12.919 --> 00:04:13.400 They did. 94 00:04:13.919 --> 00:04:18.160 It worked beautifully. It saved lives. Men stopped losing their 95 00:04:18.199 --> 00:04:19.639 teeth and dying at sea. 96 00:04:19.839 --> 00:04:22.399 Okay, so if it cured the disease and saved the fleets, 97 00:04:22.600 --> 00:04:24.160 why would they ever stop using it? 98 00:04:24.279 --> 00:04:28.319 Well, because they didn't understand the underlying chemical mechanism. They 99 00:04:28.399 --> 00:04:31.560 knew eating lemons worked, but they didn't know why. No, 100 00:04:32.319 --> 00:04:36.040 So decades later, the British Navy switched from lemons to limes. 101 00:04:36.399 --> 00:04:39.160 Limes were cheaper and easier to source from their own colonies. 102 00:04:39.199 --> 00:04:40.519 But limes aren't the same. 103 00:04:40.399 --> 00:04:44.680 Right, Limes have a significantly lower vitamin C content than lemons, 104 00:04:45.040 --> 00:04:48.519 So the scurvy came back. Oh no Yeah, people started 105 00:04:48.600 --> 00:04:52.879 dying again and the cure was essentially forgotten until ascorbic 106 00:04:52.959 --> 00:04:56.360 acid vitamin C was formally discovered in the nineteen thirties. 107 00:04:56.560 --> 00:04:58.680 That is incredible. I mean, human progress isn't just a 108 00:04:58.680 --> 00:05:01.560 straight line going up all It can invent something amazing, 109 00:05:01.800 --> 00:05:04.879 totally misunderstand the mechanism of why it works, and then 110 00:05:05.000 --> 00:05:07.680 just abandon it for an inferior tool because it's i 111 00:05:07.680 --> 00:05:08.839 don't know, cheaper or a trend. 112 00:05:08.879 --> 00:05:12.879 Ear and technological progress in computer science suffers from the 113 00:05:12.920 --> 00:05:17.839 exact same amnesia. The authors compare relying on dynamically typed 114 00:05:17.920 --> 00:05:22.720 languages like Python or JavaScript to navigating the oceans without lemons. 115 00:05:23.199 --> 00:05:24.600 Okay, break that down for me. 116 00:05:24.720 --> 00:05:27.920 So, in a dynamically typed language, the system doesn't strictly 117 00:05:28.040 --> 00:05:30.480 enforce what kind of data is flowing through it. Okay, 118 00:05:30.560 --> 00:05:33.319 that forces the programmer to hold all the metadata, all 119 00:05:33.360 --> 00:05:36.639 the rules about what kind of data goes where inside 120 00:05:36.639 --> 00:05:37.120 their own. 121 00:05:37.000 --> 00:05:41.079 Head, and human working memory is remarkably fragile. I mean, 122 00:05:41.319 --> 00:05:43.560 we can only hold on to maybe four to seven 123 00:05:43.639 --> 00:05:45.319 distinct items at any given. 124 00:05:45.079 --> 00:05:48.199 Time, exactly. So if you force a programmer to remember 125 00:05:48.240 --> 00:05:51.480 that variable x is a string of text, and variable 126 00:05:51.639 --> 00:05:52.839 y is and integer. 127 00:05:52.560 --> 00:05:55.240 And variable z is a list of user IDs. 128 00:05:55.199 --> 00:05:56.639 Right, eventually they just drop a ball. 129 00:05:56.720 --> 00:05:57.839 Yeah, you're gonna mess that up. 130 00:05:57.920 --> 00:06:00.920 Chris actually shares a painful anecdote in the text about 131 00:06:00.920 --> 00:06:05.040 this exact limitation. He once spent over two hours trying 132 00:06:05.079 --> 00:06:07.800 to track down a single type error in a dynamically 133 00:06:07.839 --> 00:06:08.680 typed language. 134 00:06:08.720 --> 00:06:09.920 Two hours for one error. 135 00:06:10.000 --> 00:06:12.240 Yeah, and the entire program was only two hundred and 136 00:06:12.240 --> 00:06:13.000 fifty lines long. 137 00:06:13.079 --> 00:06:14.199 Oh wow, that is nothing. 138 00:06:14.560 --> 00:06:17.800 But because the system allowed malformed data to just keep 139 00:06:17.879 --> 00:06:22.480 flowing downward. The error propagated far away from the original problem. 140 00:06:22.560 --> 00:06:26.120 It's like a disease spreading silently through the code exactly. So, 141 00:06:26.240 --> 00:06:29.720 dynamic typing is like navigating without lemons because we rely 142 00:06:29.800 --> 00:06:32.639 on our own limited brains to track the data types. 143 00:06:33.279 --> 00:06:36.040 How does Haskell actually solve that? What is it? Vitamin C? 144 00:06:36.360 --> 00:06:40.680 Haskell uses strict static typing. The compiler, which is the 145 00:06:40.680 --> 00:06:44.120 program that translates your code into machine language. It acts 146 00:06:44.160 --> 00:06:47.480 as an impenetrable wall before the program is even allowed. 147 00:06:47.160 --> 00:06:49.079 To run, so it checks everything. 148 00:06:48.759 --> 00:06:52.160 First, every single piece of data, against strict roles. If 149 00:06:52.160 --> 00:06:54.319 you try to hand a text string to a math function, 150 00:06:54.680 --> 00:06:57.800 the compiler stops you immediately and says this is inconsistent. 151 00:06:57.959 --> 00:06:59.839 It tracks the metadata so your brain doesn't. 152 00:06:59.600 --> 00:07:02.800 Have to You can refactor or change thousands of lines 153 00:07:02.800 --> 00:07:05.600 of code without fear knowing the compilers doing the heavy 154 00:07:05.639 --> 00:07:07.279 lifting of remembering what everything is. 155 00:07:07.759 --> 00:07:10.959 That brings us to a really crucial pivot, because to 156 00:07:11.000 --> 00:07:14.759 build a tool that can flawlessly protect our working memory, 157 00:07:15.199 --> 00:07:19.959 we need a solid mathematical foundation. And the book doesn't 158 00:07:20.000 --> 00:07:22.839 start by having you type code into a terminal. It 159 00:07:22.879 --> 00:07:25.560 actually takes us back to the nineteen thirties. 160 00:07:25.279 --> 00:07:27.839 Right to a mathematician named Alonzo Church. 161 00:07:28.160 --> 00:07:31.959 Yes, and something called the lambda calculus, which I have 162 00:07:32.040 --> 00:07:34.360 to admit. When I was first reading the sources, lambda 163 00:07:34.399 --> 00:07:36.639 calculus sounded absolutely terrifying. 164 00:07:36.759 --> 00:07:39.160 It sounds like advanced calculus, which scares a lot of. 165 00:07:39.079 --> 00:07:43.000 People, definitely, But as I dug into the actual mechanics 166 00:07:43.000 --> 00:07:46.839 of it, I realized it's really just a system of substitutions. 167 00:07:47.079 --> 00:07:49.319 It reminded me exactly of playing mad lips. 168 00:07:49.439 --> 00:07:51.120 Oh, that's a great way to think about it. 169 00:07:51.199 --> 00:07:53.360 You know those fill in the blank puzzle books where 170 00:07:53.360 --> 00:07:55.000 you ask a friend for a noun and then you 171 00:07:55.040 --> 00:07:57.040 write it into a sentence to make a funny story. 172 00:07:57.319 --> 00:08:00.839 That is a remarkably accurate way to visualize the mechanism. 173 00:08:01.720 --> 00:08:04.399 In Lenda calculus, an abstraction, which is what they call 174 00:08:04.439 --> 00:08:08.879 a function, operates just like a mad Libs sentence. Okay, 175 00:08:08.959 --> 00:08:10.920 the function has a head and a body. 176 00:08:11.160 --> 00:08:13.319 So the head of the lambda function is like the 177 00:08:13.439 --> 00:08:15.879 prompt in mad Libs. Right, it's the blank space that 178 00:08:15.920 --> 00:08:17.560 tells you what kind of word you need. It says, 179 00:08:17.800 --> 00:08:18.639 give me a noun. 180 00:08:19.560 --> 00:08:23.120 That's the parameter, right exactly, and then the body of 181 00:08:23.160 --> 00:08:25.079 the function is the actual sentence you were going to 182 00:08:25.120 --> 00:08:28.879 write that noun into. Building on your analogy, A formal 183 00:08:28.959 --> 00:08:34.320 lambda term relies on three components expressions, variables, and those abstractions. 184 00:08:35.039 --> 00:08:37.279 And a foundational rule of this system is that the 185 00:08:37.360 --> 00:08:40.600 label on that blank space doesn't matter at all. 186 00:08:40.519 --> 00:08:42.840 Meaning it doesn't matter if I ask my friend for 187 00:08:42.960 --> 00:08:46.519 no one noun a or just a describing word, right, 188 00:08:46.600 --> 00:08:48.320 as long as they give me a noun, it goes 189 00:08:48.320 --> 00:08:50.000 into the same blank space in the sentence. 190 00:08:50.320 --> 00:08:53.279 The formal mathematical term for that is alpha equivalents. It 191 00:08:53.320 --> 00:08:56.279 dictates that a function written as say, lambda x dot 192 00:08:56.480 --> 00:08:59.480 x is logically identical to a function written as lambda 193 00:08:59.639 --> 00:08:59.960 y dot com. 194 00:09:00.200 --> 00:09:01.840 Why so letters don't matter. 195 00:09:01.919 --> 00:09:04.519 The letters x and R are just placeholders. They hold 196 00:09:04.559 --> 00:09:07.159 no inherent meaning until you actually apply an argument to them. 197 00:09:07.240 --> 00:09:09.080 Okay, so we have our mad lib sentence, We have 198 00:09:09.080 --> 00:09:10.879 our blank space, and ask my friend for a noun 199 00:09:10.960 --> 00:09:12.559 and they say banana. 200 00:09:12.720 --> 00:09:14.200 Classic mad libs. 201 00:09:13.960 --> 00:09:17.279 Always, I take the word banana, I substitute it into 202 00:09:17.279 --> 00:09:19.320 the blank space in the sentence, and I cross out 203 00:09:19.360 --> 00:09:23.320 the prompt. That mechanical process of substituting and crossing out 204 00:09:23.320 --> 00:09:27.120 the prompt, that's what the text calls beta reduction, right. 205 00:09:27.120 --> 00:09:30.759 Yes, exactly. In beta reduction, you apply a function to 206 00:09:30.799 --> 00:09:34.759 an argument. You take the input expression, substitute it for 207 00:09:34.879 --> 00:09:38.399 all instances of the bound variable inside the body of the. 208 00:09:38.320 --> 00:09:40.039 Function, and then cross out the prompt. 209 00:09:40.039 --> 00:09:43.799 Crucially, yes, you eliminate the head. The head's only purpose 210 00:09:43.879 --> 00:09:46.799 was to bind the variable. Once the substitution is complete, 211 00:09:46.879 --> 00:09:47.879 it just vanishes. 212 00:09:48.240 --> 00:09:50.679 This is a massive light bulb moment for me. If 213 00:09:50.720 --> 00:09:55.799 we strip away the intimidating jargon, evaluation is really just simplification. 214 00:09:55.480 --> 00:09:59.039 It really is. The authors provide a very intuitive mathematical 215 00:09:59.080 --> 00:10:03.320 example to prove this. Consider the expression two thousand divided 216 00:10:03.360 --> 00:10:06.320 by one thousand. Okay, that is a mathematically valid expression, 217 00:10:06.399 --> 00:10:07.840 but it is not fully reduced. 218 00:10:07.919 --> 00:10:09.919 Right Because if someone asked me how many apples I had, 219 00:10:10.000 --> 00:10:12.639 I wouldn't say I have two thousand divided by one 220 00:10:12.679 --> 00:10:13.360 thousand apples. 221 00:10:13.480 --> 00:10:14.519 That would be very weird. 222 00:10:14.960 --> 00:10:16.879 My brain has to do extra work to parse that 223 00:10:16.960 --> 00:10:18.399 I have two apples. 224 00:10:18.759 --> 00:10:22.679 Exactly. The number two is the normal form of that expression. 225 00:10:23.480 --> 00:10:27.039 Beta normal form is achieved when you cannot apply any 226 00:10:27.080 --> 00:10:29.240 more lambdas to arguments. 227 00:10:29.519 --> 00:10:31.240 So you're at the end of the line. 228 00:10:31.000 --> 00:10:34.039 Right, There are no more evaluation steps to take. Nothing 229 00:10:34.080 --> 00:10:37.440 is left to simplify the ultimate goal of any program 230 00:10:37.480 --> 00:10:40.360 in this paradigm is to reach its normal form, where 231 00:10:40.360 --> 00:10:42.399 it is completely evaluated. 232 00:10:42.840 --> 00:10:46.320 In tying this back to our overarching theme, simplifying down 233 00:10:46.399 --> 00:10:49.200 to a normal form is another way to save cognitive load. 234 00:10:49.320 --> 00:10:49.480 Oh. 235 00:10:49.519 --> 00:10:52.519 Absolutely, you aren't forcing the programmer or the machine to 236 00:10:52.600 --> 00:10:55.559 juggle unresolved equations. You distill it down to its most 237 00:10:55.679 --> 00:10:58.919 basic truth. But I guess how does this nineteen thirties 238 00:10:59.000 --> 00:11:01.840 math translate into actual software engineering. 239 00:11:02.039 --> 00:11:05.159 Well, it translates through the concept of pure functions. Because 240 00:11:05.159 --> 00:11:08.240 Haskell is built on lamb to calculus, it relies entirely 241 00:11:08.279 --> 00:11:11.279 on these pure mathematical relations. Okay, there is a quote 242 00:11:11.279 --> 00:11:13.799 from Mike Hammond in the text that captures the essence 243 00:11:13.840 --> 00:11:17.279 of this perfectly. He says, functions are beacons of constancy 244 00:11:17.320 --> 00:11:18.919 in a sea of turmoil. 245 00:11:18.919 --> 00:11:21.200 Beacons of constancy. I like that. But what makes a 246 00:11:21.200 --> 00:11:23.480 function pure or constant in a machine? 247 00:11:23.720 --> 00:11:27.720 It comes down to a property called referential transparency. A 248 00:11:27.799 --> 00:11:31.960 pure function is a strict, unbreakable relation between inputs and outputs, 249 00:11:32.039 --> 00:11:34.879 so it's locked in completely. If a function is called 250 00:11:35.120 --> 00:11:37.399 F and you give it the input, one and it 251 00:11:37.440 --> 00:11:40.960 returns the output A. It must return A every single 252 00:11:41.000 --> 00:11:43.039 time it is evaluated, no matter. 253 00:11:42.759 --> 00:11:44.799 What time of the day the computer is running, no 254 00:11:44.840 --> 00:11:46.320 matter what other apps. 255 00:11:46.039 --> 00:11:48.600 Are open, no matter what is happening on the network. Yes, 256 00:11:48.919 --> 00:11:52.360 if a function returns X on Tuesday and then returns 257 00:11:52.519 --> 00:11:55.879 why on Wednesday for the exact same input, it is 258 00:11:55.919 --> 00:11:58.279 an invalid function in functional programming. 259 00:11:58.440 --> 00:12:00.519 But doesn't that happen a lot in other languages. 260 00:12:00.639 --> 00:12:04.240 It happens all the time in dynamically typed languages, because 261 00:12:04.279 --> 00:12:06.919 a function might secretly rely on a global variable that 262 00:12:07.080 --> 00:12:08.840 changed somewhere else in the system. 263 00:12:09.039 --> 00:12:09.720 Ah, I see. 264 00:12:09.720 --> 00:12:12.159 That is called a side effect. Pure functions have no 265 00:12:12.240 --> 00:12:12.879 side effects. 266 00:12:12.960 --> 00:12:15.799 This is exactly how we prevent the working memory overlow. 267 00:12:15.840 --> 00:12:18.159 We talked about what the scurvy analogy. If you know 268 00:12:18.240 --> 00:12:21.120 for an absolute fact that an input always yields the 269 00:12:21.200 --> 00:12:23.919 exact same output, you don't have to keep a mental 270 00:12:23.919 --> 00:12:26.960 list of exceptions. You don't have to wonder, wait, did 271 00:12:27.000 --> 00:12:29.200 this variable change while I wasn't looking right? 272 00:12:29.240 --> 00:12:32.320 You can trust the bricks you're building with That predictability 273 00:12:32.360 --> 00:12:37.200 allows you to build massive, sprawling software systems out of small, reliable, 274 00:12:37.279 --> 00:12:38.399 independent functions. 275 00:12:38.440 --> 00:12:40.919 You just assemble them like interlocking gears. 276 00:12:40.679 --> 00:12:44.480 Exactly knowing they will never spontaneously change shape or turn 277 00:12:44.480 --> 00:12:45.320 in the wrong direction. 278 00:12:45.519 --> 00:12:48.879 Because these pure functions are perfectly predictable because they never 279 00:12:49.000 --> 00:12:51.879 change based on hidden variables, you don't need to boot 280 00:12:51.960 --> 00:12:54.960 up a massive, complex program just to see if one 281 00:12:55.039 --> 00:12:56.159 tiny piece works. 282 00:12:56.559 --> 00:12:57.039 You don't. 283 00:12:57.320 --> 00:12:59.559 You can just ask the machine to evaluate a single 284 00:12:59.559 --> 00:13:02.960 function in isolation, which brings us to how we actually 285 00:13:02.960 --> 00:13:05.399 sit down and talk to this system on a computer screen. 286 00:13:05.559 --> 00:13:08.759 Right getting to the practical side, the sources introduced two 287 00:13:08.919 --> 00:13:11.600 main ways to interact with Haskell. You can write your 288 00:13:11.679 --> 00:13:14.159 code into permanent source files, or you can use a 289 00:13:14.320 --> 00:13:15.360 r EPL. 290 00:13:15.320 --> 00:13:17.559 R EPL read evil print loop. 291 00:13:17.600 --> 00:13:22.720 Exactly in Askel's case. This interactive environment is called GHCI. 292 00:13:23.360 --> 00:13:25.559 It is a prompt where you type in an expression, 293 00:13:25.879 --> 00:13:29.720 hit enter, and the system instantly reads it, evaluates it 294 00:13:29.759 --> 00:13:32.360 to its normal form, and prints the result back to you. 295 00:13:32.679 --> 00:13:34.840 So it's a laboratory for immediate feedback. 296 00:13:34.879 --> 00:13:35.840 That's a great way to put it. 297 00:13:35.879 --> 00:13:38.919 But the way we type things out, the syntax has 298 00:13:38.960 --> 00:13:41.720 some quirky rules that can really trip up a beginner, 299 00:13:42.000 --> 00:13:44.159 particularly how we order the words and symbols. 300 00:13:44.279 --> 00:13:48.320 Yeah, this raises the distinction between prefix and infix functions. 301 00:13:48.840 --> 00:13:53.519 By default, functions in Haskell use prefixed syntax. The function 302 00:13:53.639 --> 00:13:56.519 name comes first before the arguments. So if you have 303 00:13:56.559 --> 00:13:58.799 a function that triples a number, you type the word 304 00:13:58.799 --> 00:14:00.799 triple a space and then the number. 305 00:14:00.559 --> 00:14:03.480 Two, soly issuing a command in English triple this number 306 00:14:03.480 --> 00:14:04.440 two exactly. 307 00:14:04.840 --> 00:14:08.399 But math operators like the plus sign or the multiplication 308 00:14:08.519 --> 00:14:12.120 sign are actually functions too. They take inputs and return 309 00:14:12.120 --> 00:14:15.480 an output. However, they are infix operators, so they sit. 310 00:14:15.360 --> 00:14:17.600 In the middle of their arguments like one plus one. 311 00:14:17.720 --> 00:14:19.559 Right, which makes sense because that's how we all learn 312 00:14:19.600 --> 00:14:22.080 math in elementary school. But the book points out that 313 00:14:22.120 --> 00:14:25.600 mixing these up introduces strict rules about associativity. 314 00:14:26.000 --> 00:14:28.440 Hold on, if you're listening to this while commuting or 315 00:14:28.480 --> 00:14:31.279 making dinner, your eyes might have just glazed over at 316 00:14:31.320 --> 00:14:34.480 the word associativity. Let's ground this fair point. Why does 317 00:14:34.519 --> 00:14:36.759 a programmer actually need to care if the math goes 318 00:14:36.840 --> 00:14:38.759 left to right or right to left. 319 00:14:38.720 --> 00:14:41.519 Because the computer will follow its rules flawlessly, whether you 320 00:14:41.600 --> 00:14:44.799 intended it to or not, which can lead to wildly 321 00:14:44.840 --> 00:14:45.639 different results. 322 00:14:45.840 --> 00:14:46.679 Give me an example. 323 00:14:46.759 --> 00:14:49.480 Okay, if you type two, three four in into the 324 00:14:49.639 --> 00:14:54.360 r epl, the machine sees two exponent operators. Because exponentiation 325 00:14:54.519 --> 00:14:58.159 associates to the right, it calculates the right most part first, 326 00:14:58.320 --> 00:15:00.840 so three four right, which is e eighty one. Then 327 00:15:00.879 --> 00:15:04.279 it calculates to eighty one, which is an astronomically large number. 328 00:15:04.360 --> 00:15:06.639 Wow. But if you assumed it read left to right 329 00:15:07.159 --> 00:15:09.960 like regular English text, you would expect it to calculate 330 00:15:09.960 --> 00:15:12.440 two three first, which is eight right, and then raise 331 00:15:12.480 --> 00:15:15.279 eight to the fourth power, which is only four ninety 332 00:15:15.320 --> 00:15:16.600 six exactly. 333 00:15:16.639 --> 00:15:19.360 So if your software was calculating a financial interest rate 334 00:15:19.559 --> 00:15:23.080 and you misunderstood the associativity, you just bankrupted the company. 335 00:15:23.159 --> 00:15:26.399 Yikes. A perfect example of why I understand the strict 336 00:15:26.480 --> 00:15:29.320 rules of the language prevents catastrophic failure. 337 00:15:29.480 --> 00:15:32.919 Yeah, and the book dives into another concrete mathematical example 338 00:15:32.919 --> 00:15:36.960 that illustrates this beautifully, the modulo clock using the mod 339 00:15:37.080 --> 00:15:38.120 and REM functions. 340 00:15:38.320 --> 00:15:40.639 Both of those deal with division, but mod stands for 341 00:15:40.759 --> 00:15:43.960 modulo and REM stands for a remainder. Walk me through 342 00:15:44.080 --> 00:15:46.159 modular arithmetic using a standard clock. 343 00:15:46.320 --> 00:15:49.120 Sure, look at a standard twelve hour analog clock. 344 00:15:49.320 --> 00:15:49.600 Yeah. 345 00:15:49.639 --> 00:15:52.000 If the current time is eight zero zero and you 346 00:15:52.080 --> 00:15:53.919 need to know what time it will be in eight hours, 347 00:15:54.279 --> 00:15:56.279 you don't add eight plus a and declare it will 348 00:15:56.320 --> 00:15:57.159 be sixteen o'clock. 349 00:15:57.440 --> 00:16:01.240 Obviously not on a twelve hour dial. The numbers wrap 350 00:16:01.279 --> 00:16:04.200 around the modulus, which is twelve, So you add four 351 00:16:04.240 --> 00:16:07.519 hours to reach twelve zero zero. The dial resets to 352 00:16:07.600 --> 00:16:10.279 zero and you add the remaining four hours. So eight 353 00:16:10.320 --> 00:16:12.960 hours after eight zero zero is four point zero. 354 00:16:13.120 --> 00:16:16.879 That is arithmetic modulu twelve. Haskell actually includes two different 355 00:16:16.919 --> 00:16:19.120 functions for this, mod and REM, and they usually the 356 00:16:19.159 --> 00:16:23.399 exact same answer, usually, but they diverge in behavior when 357 00:16:23.399 --> 00:16:27.799 you introduce negative numbers. If one argument is negative, mod 358 00:16:28.000 --> 00:16:30.120 takes the sign of the divisor the number you are 359 00:16:30.120 --> 00:16:32.759 dividing by, okay, but REM takes the sign of the 360 00:16:32.759 --> 00:16:34.360 dividend the number being divided. 361 00:16:34.440 --> 00:16:36.360 Okay. Let's put this in a real world scenario so 362 00:16:36.399 --> 00:16:38.879 we understand why this matters. Let's say you are building 363 00:16:38.919 --> 00:16:39.960 a calendar application. 364 00:16:40.080 --> 00:16:41.360 Okay, a calendar app. 365 00:16:41.360 --> 00:16:43.960 Days of the week loop every seven days, so your 366 00:16:44.000 --> 00:16:48.440 modulus is seven. Tuesday is day two. A user wants 367 00:16:48.480 --> 00:16:50.720 to search for an event that happened ten days ago. 368 00:16:51.159 --> 00:16:54.159 You subtract ten from two, which gives you negative eight. 369 00:16:54.759 --> 00:16:57.000 Right, and if you use the REM function to divide 370 00:16:57.039 --> 00:16:59.600 negative eight by seven, the remainder takes the sign of 371 00:16:59.600 --> 00:17:02.080 the negative of eight. It returns in negative. 372 00:17:01.759 --> 00:17:03.960 One, but there is no negative first day of the week. 373 00:17:04.000 --> 00:17:05.960 The calendar application completely. 374 00:17:05.519 --> 00:17:08.480 Breaks exactly, but if you use the mod function, it 375 00:17:08.559 --> 00:17:11.720 correctly wraps around the negative value and returns a positive six, 376 00:17:11.759 --> 00:17:13.119 which corresponds to Saturday. 377 00:17:13.240 --> 00:17:13.640 Wow. 378 00:17:14.039 --> 00:17:17.720 Knowing that subtle distinction prevents massive bugs when you are 379 00:17:17.720 --> 00:17:21.480 dealing with wrapped values like times, dates, or even graphical 380 00:17:21.559 --> 00:17:22.480 user interfaces. 381 00:17:22.839 --> 00:17:25.960 But thinking about typing all of this outkeeping track of 382 00:17:26.519 --> 00:17:31.319 prefix infix associativity whether a value might be negative, I mean, 383 00:17:31.440 --> 00:17:34.599 it feels like the code could become incredibly dense and 384 00:17:34.759 --> 00:17:36.400 visually overwhelming to read. 385 00:17:37.000 --> 00:17:40.519 It could, which is why programming languages utilize something called 386 00:17:40.599 --> 00:17:41.680 syntactic sugar. 387 00:17:42.079 --> 00:17:47.000 Syntactic sugar beyond sounding delicious, what is the actual mechanism here? 388 00:17:47.200 --> 00:17:50.559 Syntactic sugar is a feature of a programming language's grammar 389 00:17:50.640 --> 00:17:53.400 designed strictly to make things easier for humans to read 390 00:17:53.440 --> 00:17:56.119 and write. Oh nice, Yeah, It does not change the 391 00:17:56.200 --> 00:18:00.000 underlying semantics or the lambda calculus math at all. For example, 392 00:18:00.000 --> 00:18:03.279 simple to make a number negative in Haskell, the formal 393 00:18:03.359 --> 00:18:06.599 mechanism is to apply a function called negate to it, 394 00:18:06.880 --> 00:18:07.480 so you'd. 395 00:18:07.240 --> 00:18:09.000 Have to type the word negate, add a space, and 396 00:18:09.039 --> 00:18:10.319 tack the number nine right. 397 00:18:10.559 --> 00:18:13.400 But seeing the word negate everywhere clutters the screen, So 398 00:18:13.519 --> 00:18:15.440 Haskell allows you to just type negative nine. 399 00:18:15.559 --> 00:18:18.839 Ah. So behind the scenes, the compiler is still translating 400 00:18:18.839 --> 00:18:22.279 that little dash into the formal negate function. Hmm, but 401 00:18:22.319 --> 00:18:24.799