It's really cool to watch your videos, because they open our minds to other possibilities of how to build the same code. You teach things that are not easy to find anywhere else.
Little Vim tip: full stop (.) repeats your last action. This includes entering insert mode and typing things. e.g. at 6:10 you type out "_avg" twice. You could instead just type it once, navigate to the next function, then press . and that'll type out "_avg" again for you. Same with _minmax a few seconds later! Edit: Just watched the part about partial functions. Can't believe I've never seen them before, they seem like such a useful tool!
But also in vscode you can edit hundreds of lines at once (and I have done so in a productive way before)... I take people who talk about vim/emacs being best in the same way I take people talking up the King James Version. Sure, it's the Bible, but the language is archaic and outdated. Vim is great, but so is vscode. And this is coming from someone who is efficient in both. I definitely prefer vscode.
People think Emacs is old, but they forget that it was built by a whole bunch of smart hackers who lived and breathed coding, knew what they were doing, and were not beholden to marketing managers at any company. For example, they used an advanced programming language (LISP), which can still scare the pants off the Java/C♯/PHP crowd today. Does the automation language in _your_ favourite editor understand closures and lexical binding? (Interestingly, the sort of concepts discussed in this very video.)
I like Vim too but it's not good at bigger refactors. So I use PyCharm for bigger projects, and sometimes I just switch on the Vim plugin to do some stuffs. For smaller things (indenting the whole file, removing indentation, adding comma to the and of lines, join all the lines, removing every fifth lines with macro) I just start a Vim in command line.
This came in handy for my weekend project! I’ve watched through all your Python videos so far, and my code has started to improve dramatically as a result. Thanks for sharing!
wow. I am currently working on a project where I use functions that return customized functions, just as in your example. So partial just blew my mind a little. Thank you so much for introducing this awesome tool to me :)
I guess it matters how you define the function of the bread. I may be partial to German bread and will happily eat it by itself, but there's an argument to be made that, say, certain French breads shine as part of a breakfast. With American bread, islice it and toast it and it slaps. Italian bread is in a class of its own. I haven't tried Dutch bread, I gotta find some, I wonder if they import it here. And to give y'all some closure, I'm sure any bread from any country can be great if combined with the right ingredients
Dutch bread, it’s simply a slice of very soft very limp toast called Roti … maybe that’s why a Dutch can’t understand other countries being fond of their bread … I’ve been living in Indonesia and they had adopted Dutch bread-making habits, while Vietnam for example offer baguette due to the French, and none of them compares to German variety of course …. 😜
Great video, Arjan. I also use the functools a lot, e.g. partial. What I really love about Python is that it supports many different approaches and does not force you to use a single one. As a developer I can choose between many different approaches, and choose the one that I think is the best to solve my current problem. Python lets me express myself in code directly in various ways.
man your channel is underrated!! you are really provide a high quality contents with a great knowledge and experience, partial function are awesome i need to use it more, thank you.
I really like how your tutorials have never been completely OOPs. A functional-leaning python programmer is a rarity :D I wonder if you have a playlist of videos on functional-ish concepts. The way you describe things is perfectly applicable and would be a very clean and organised way to write julia code as well.
I wouldn't expect a lot of those types of videos due to Python's fairly limited support of functional programming. A few functional modules that are nice to use in the standard library (which might be covered later on) are itertools, operator, and the map, any, all, and filter functions.
@@DerekHohls Yes. I think this tends to get understated -- I've done a lot of Python and PowerShell "programming" for a lot of real-world enterprise uses, and I very, very rarely have to use an OOP approach when coding something from scratch. I rarely have to define my own classes at all; I wouldn't say most/all/anything superlative, but there's a lot of people who understand OOP but just never need to use it Python in that way.
Hi Arjan! Thanks for all your work teaching and coding! I came accross partial when I was working with python's multiprocessing library for performing a task with many files with some "standard" parameters that were the same for each and some "variable" parameters (such as the dataframe) that were unique to each process. This video gave me a better context of what partial actually is/does, and it's helping me with an NLP project refactor. 😄
functools is amazing. Besides partial there is also reduce, wraps and lru_cache which I use all the time. singledispatch can also be nice in very simple cases but it quickly hits a roadblock in more complex scenarios.
I've temporarily moved to Finland from the Netherlands, and I can confirm: the bread is better in the Netherlands. There is a even bigger difference in cheese though: the cheese is horrible here.
This is one of your best videos yet (and they're all consistently great!) I had to post a comment just to give extra likes 👍👍👍👍👍 The bits about closures and partial are BOSS 🤓
Hi Arjan, Your content is great, found that one amusing in particular, since my wife is Dutch living in Israel for almost 20 years, while still complaining over the local bread taste and praising the Dutch version 🤣hilarious !!! Thanks for the effort you put into the content. 🙏
Great video as always! Partials are nice. I wrote a validate(validate_func, data_type, data, ....) function that I could then call from a partial like validate_json = partial( validate, json.loads, "JSON") or validate_uuid = partial(validate, uuid.UUID4, "UUID").
you told me to buy Dutch bread after getting the free guide, but I see no link in video description for where to buy it, so I guess I will just keep living in my "bread fantasy world". Love your vids. They are always very informative. Thank you!
get_best_bread = partial(get_bread_from, country_code="de_DE") :) Great video! I was astounded seeing the closure part and then really impressed seing partial() in use. The flexibility and simplicity of providing values for named variables and using it on any function is amazing. My wife is gonna get jealous of the love I am developping for Python. Thanks for sharing this!
Thank you for another great video Arjan! also, your anecdote reminded me of a similar pun "no matter how kind your kids are, German kids will always be kinder"
Is there any difference between using a partial function and a lambda function? For some f(x,y), shouldn't lambda x: f(x, y0) be equivalent to partial(f, y=y0) ?
While I don't follow everything you do, I like to watch your videos because it exposes me to new ways of thinking as well as showing me the practical reality of what is typed to accomplish an objective. Very cool indeed. :)
Very informative! As someone who was previously unaware of partials, I have achieved similar functionality by setting constants and using them as default parameters (as you mentioned). This approach seems cleaner, though, so I will definitely give it a whirl in the future!
12:35 But you could replace it with a docstring. Which could be dynamically generated, to incorporate the actual limit, and assigned to the ＿＿doc＿＿ property of the function object being returned.
Nothing really related to your video, but I wanted to thank you for your work. Not only it is very well explained but it really inspires me by debunking some dev "features" and mindset related to development. I don't really study your videos (sometimes still too advanced for me), but they show me every time the beauty of code and keeps me learning further. Furthermore, I tend to really enjoy your both light/humble and technical tone, I'm from marketing, and your channel is doing marketing right focusing on content value first. So again, thank you for your work :)
Just yesterday, it inspired me to have fun playing with inheritance and composition with PyMongo's source code (not touching at it, but making my way around it for some purpose). Trying to deep dive into code like you do, finding beautiful and fun features, far from just focused tech tutorials, more of a technical case study / learning experience. thx !
Do partials have any affect on time complexity of a function? Say you have a function where the part that is partially applied takes O(n^2), and the remaining part would be O(n). When calling the partially applied function, would this be O(n), because the O(n^2) part has already been handled or O(n^2) because it hasn't? This seems quite niche, but im currently working on graphical demonstration of time complexities and partials could be very helpful as I could only look at timing a specific part of a function. I watched this video a while back and remembered how useful I found it
Someone probably mentioned this already, but if you refactor the functions you're passing to partial so the arguments you're specifying therein are to the left of the one the new function still needs, you don't need to pass them by keyword.
@@martinleonhardt1541 Partial will work around mutable types and closure better. Lambda got me in a pickle when I tried using it in a list comprehension.
I constantly write code that starts dead simple and expands in options progressively, ending up with very customizable functions. I’m gonna use partials to take care of this now!
You said 35_000_00 is thirty five thousand, but you also said that the underscores doesn't change anything about the integer value so it would be 3500000 so that wouldn't add 2 decimals randomly? So it should be 3_500_000 which is 3,5 million?
How would you do dependency injection for functions? For instance I have several functions where I want to use a database, for this I have a class or function which I want to re-use and mock for unit tests. If I use classes I would inject this dependency in the constructor. In the composition root I would assemble my objects. If I would do the same on function level for instance with te partial() it sounds like a lot of configuration, and a bit unclear for me how to structure the code then.
@@ArjanCodes I am a developer from Taiwan( in Asia). I have been writing Scala code as a data engineer for the last three years. Recently I joined a new company and my team uses Python for everything. And I struggle to accustom myself to the new environment. Watching your videos about Python’s usage and how to write code in a more functional programming way makes me more confident and more productive. Thanks for the effort!
Why did he put the extra 2 zeros on 35_000_00 ? It evaluates in my interpreter as 3,500,000 when I'm sure he meant 35,000. Could someone please explain what I'm missing?
French bread is of course better by far than any other bread, but I guess it wouldn’t be fair to include it in the competition as there wouldn’t be any controversy.
My mother used to make bread with recipes that undoubtedly came from the Netherlands (as my country was a Dutch colony and before that we have no bread) and I'll say they're great! 👍🏽
I rarely comment on video tutorials, unless it gave me that "aha" moment, and this was one of those, thanks ArjanCodes for educating us! btw, have you profile which "way" is more faster, the OOP or FP style?
I have studied industrial engineering in Spain, with little curriculum in programming. Is it possible that I prefer functional programming due to my background in engineering? I.e. decompose the 'problem' in little steps via functions and then there is a main function that calls functions which in turn call lower level functions...and so on.
While I use functools in many occasions I always wondered what the (technical) difference is between the partial function and creating a lambda (or even a function) simply reducing the arguments? I mean, i could also do buy_strategy = lambda prices: should_buy_average(prices, 4) While "partial" usually increases readability I wondered if there are any differences in terms of performance.
@@marsma18 kind of, yes, but not in the way he showed closures in the video as my approach IS a function with a valid signature actually calling another function while the approach in the video RETURNS a newly generated function with a valid signature. Moreover I was interested in the technical difference than the logical difference.
@@comedyclub333 The only real difference that I came across was while implementing a naive save system in my game engine (naive in the sense that I just dump whole objects, including attached callback functions and everything to file using pickle). Now callback functions cannot typically have arguments, but often require a variety of them, so I end up using functools.partial fairly often - because lambdas are anonymous and cannot be pickled, while partial functions can.
@@RitchieDiamond That's actually something really interesting. I just tried it and you are right! It seems that lambdas really are not savable into serialized data. Crazy!
Very useful if you want to write to a file using print: with open("output.txt", "w") as f: fprint = partial(print, file=f) fprint("Hello world!") Same for printing to stderr import sys eprint = partial(print, file=sys.stderr) ... eprint(f"Error occurred: {error.description}")
Even if I'm not yet at the level where I have to implement most of the things you share, I still go ahead to watch your videos anyway. You teach well. That said, I thought I should point out something to you (and to those not familiar with this style). At minute 12:51, you add two additional zeroes, presumably to make the number 35,000. But that actually changes the number to 3,500,000. (And I confirmed this on my python shell.) 35_000_00 != 35,000 Instead, 35_000_00 == 3,500,000 35_000 == 35,000 3_4 == 34 (just to show those who might not know that you don't always have to end with zeroes) In essence, it seems to me you think adding two zeroes after the last underscore means python will treat the variable as a float. It won't. It'll just add two more zeroes to the integer. If you want a floating point number: 35_000.00 1_2.3 12_34_567_8009.027 are all different valid examples.
Money is best dealt with in integers because floats will have errors dividing in binary. It is 35000.00. When returning values, you just have to deal with the extra 00 as cents.
Hi, Your videos are awesome... one question though. How do you place that run button beside the split editor right icon in vs code, are you using any plugins/extensions?
I'm a bit confused. in your dataclass you are defining buy_strategy as a TradingStrategyFunction which you defined as that Callable.. But what's the difference between that and just calling the definition directly from within the class? Without self.function_name() just function_name()?
The reason I'm using a function and not a method, is that by defining them as functions, you can set them when you create the object (this is what I do in the main function). If the buy/sell strategies were methods, you would need to create a subclass for each combination of buy/sell strategy you want, so it's less flexible.
Nice video as usual! Question on partial. Before knowing about it, I used to do: buy_strategy = lambda prices: should_buy_average(prices, window_size=4) vs buy_strategy = partial(should_buy_average, window_size = 4) I think the result is the same. Are these methods equivalent, but partial with less code?
There are a few differences I guess. For example with partial you can change the should_buy_average function (as long as it still takes the window_size) and need not change the partial statement, whereas you would need to change the lambda statement because it explicitly uses the parameters the should_buy_average function expects. You can also change the parameters' order of the function and the partial application still works.
Hi Arjan, Please continue the Design Pattern Series, You can take one by one different design patterns with practical code for python. There is not much good content out in youtube.
I'm interested in what keys you're pressing. How do you: 6:06 forward-delete an argument including the following comma and space up to the next argument? 15:31 delete a whole line at a time incl. LF? 15:49 delete to the end of the line? 17:59 forward-delete up to the end of the argument list? Also British people know that our bread is the worst so we don't even pretend it's good.
Hi Arjan. Thank you for the instructions on passing extra params using closures. I bookmarked this video when I first watched it because I suspected that it would be useful, and today it was exactly what I needed to add functionality to the callback function passed to ftplib.retrbinary()
10:38 I like to use a name like “def_should_buy_avg”. The extra “def” on the front looks like the keyword for defining a function -- which is what the function does.
Functional patterns are awesome and one of my favourite ways to structure code, u see it a lot in scala or heaven forbid R (their class like defns are terribad) But scalas map, filter, reduce, foldleft, currying or partial hell maybe i just used clojures and they got tail recursive. I had all of my steps be like functional but isolated environments be classes. I think both are cool
I just discovered a disappointing downside of functools partial - partially applied functions don't typecheck! I use mypy to check my type hints. I have a type hint for a "validate_multiple" function that expects a list of Callables with a certain validator signature. I tried to pass a partial to it and got this error: Argument 3 to "validate_multiple" has incompatible type "List[object]"; expected "Sequence[Callable[[Dict[Any, Any], ExecutionContext], Dict[Any, Any]]] Apparently there's an ugly workaround that involves manually casting the partial back to the right type, but that's not satisfying at all. :(
Forget where I saw it on youtube, but someone commented that most classes are basically "collections of partially applied functions". In that scenario where I have similar methods, I would rather have user instantiate a class once than do the partial applications repetitively for the different functions / methods
It's nice that you can replace interfaces and their implementations with partial functions. However I'm still concerned with what happens in cases where we need to do some error validation and type checking, any ideas?
💡 Get my FREE 7-step guide to help you consistently design great software: arjancodes.com/designguide.
It's really cool to watch your videos, because they open our minds to other possibilities of how to build the same code. You teach things that are not easy to find anywhere else.
You sir, are putting out the best Python content on the internet. Thank you!
Little Vim tip: full stop (.) repeats your last action. This includes entering insert mode and typing things. e.g. at 6:10 you type out "_avg" twice. You could instead just type it once, navigate to the next function, then press . and that'll type out "_avg" again for you. Same with _minmax a few seconds later!
Edit: Just watched the part about partial functions. Can't believe I've never seen them before, they seem like such a useful tool!
You’re absolutely right. I still have a lot to unlearn from not editing with Vim for about 35 years.
Wow. I really need to learn Vim properly and use it.
But also in vscode you can edit hundreds of lines at once (and I have done so in a productive way before)... I take people who talk about vim/emacs being best in the same way I take people talking up the King James Version. Sure, it's the Bible, but the language is archaic and outdated.
Vim is great, but so is vscode. And this is coming from someone who is efficient in both. I definitely prefer vscode.
People think Emacs is old, but they forget that it was built by a whole bunch of smart hackers who lived and breathed coding, knew what they were doing, and were not beholden to marketing managers at any company. For example, they used an advanced programming language (LISP), which can still scare the pants off the Java/C♯/PHP crowd today. Does the automation language in _your_ favourite editor understand closures and lexical binding? (Interestingly, the sort of concepts discussed in this very video.)
I like Vim too but it's not good at bigger refactors. So I use PyCharm for bigger projects, and sometimes I just switch on the Vim plugin to do some stuffs. For smaller things (indenting the whole file, removing indentation, adding comma to the and of lines, join all the lines, removing every fifth lines with macro) I just start a Vim in command line.
This came in handy for my weekend project! I’ve watched through all your Python videos so far, and my code has started to improve dramatically as a result. Thanks for sharing!
You’re welcome - glad to hear that the videos help you improve your code.
wow. I am currently working on a project where I use functions that return customized functions, just as in your example. So partial just blew my mind a little. Thank you so much for introducing this awesome tool to me :)
I guess it matters how you define the function of the bread. I may be partial to German bread and will happily eat it by itself, but there's an argument to be made that, say, certain French breads shine as part of a breakfast. With American bread, islice it and toast it and it slaps. Italian bread is in a class of its own. I haven't tried Dutch bread, I gotta find some, I wonder if they import it here. And to give y'all some closure, I'm sure any bread from any country can be great if combined with the right ingredients
Dutch bread, it’s simply a slice of very soft very limp toast called Roti … maybe that’s why a Dutch can’t understand other countries being fond of their bread … I’ve been living in Indonesia and they had adopted Dutch bread-making habits, while Vietnam for example offer baguette due to the French, and none of them compares to German variety of course …. 😜
Great video, Arjan. I also use the functools a lot, e.g. partial. What I really love about Python is that it supports many different approaches and does not force you to use a single one. As a developer I can choose between many different approaches, and choose the one that I think is the best to solve my current problem. Python lets me express myself in code directly in various ways.
Thank you Robert! It’s indeed one of the powers (and dangers) of Python that it allows for so much flexibility.
One of the greatest videos I've seen from your TH-cam channel, Arjan! Thanks for it.
man your channel is underrated!! you are really provide a high quality contents with a great knowledge and experience, partial function are awesome i need to use it more, thank you.
Thanks Ziad, happy you’re enjoying the content and it was helpful!
I really like how your tutorials have never been completely OOPs. A functional-leaning python programmer is a rarity :D I wonder if you have a playlist of videos on functional-ish concepts. The way you describe things is perfectly applicable and would be a very clean and organised way to write julia code as well.
I wouldn't expect a lot of those types of videos due to Python's fairly limited support of functional programming. A few functional modules that are nice to use in the standard library (which might be covered later on) are itertools, operator, and the map, any, all, and filter functions.
@@siddsp02 Thanks!
Would also appreciate a treatment of the functional approach and its strengths. Also, mentioning how to write good functions in practice.
Actual of lot of folks who are not programmers but who use Python for basic data processing typically would never use a class at all.
@@DerekHohls Yes. I think this tends to get understated -- I've done a lot of Python and PowerShell "programming" for a lot of real-world enterprise uses, and I very, very rarely have to use an OOP approach when coding something from scratch. I rarely have to define my own classes at all; I wouldn't say most/all/anything superlative, but there's a lot of people who understand OOP but just never need to use it Python in that way.
Hi Arjan! Thanks for all your work teaching and coding! I came accross partial when I was working with python's multiprocessing library for performing a task with many files with some "standard" parameters that were the same for each and some "variable" parameters (such as the dataframe) that were unique to each process. This video gave me a better context of what partial actually is/does, and it's helping me with an NLP project refactor. 😄
functools is amazing. Besides partial there is also reduce, wraps and lru_cache which I use all the time. singledispatch can also be nice in very simple cases but it quickly hits a roadblock in more complex scenarios.
Thank you for your time and patient, the DOC contains a beautiful explanation in 7 steps about design better.
I've temporarily moved to Finland from the Netherlands, and I can confirm: the bread is better in the Netherlands.
There is a even bigger difference in cheese though: the cheese is horrible here.
I feel this way about cheese when moving to Norway
Finnish rye bread is unbeatable.
This is one of your best videos yet (and they're all consistently great!) I had to post a comment just to give extra likes 👍👍👍👍👍
The bits about closures and partial are BOSS 🤓
Thank you so much!
I’ve been living in the Netherlands for 3 years and you are the first dutch that I know that doesn’t like sliced bread 🤣
Very good video, as usual!
Hi Arjan,
Your content is great, found that one amusing in particular, since my wife is Dutch living in Israel for almost 20 years, while still complaining over the local bread taste and praising the Dutch version 🤣hilarious !!!
Thanks for the effort you put into the content. 🙏
Great video as always! Partials are nice. I wrote a validate(validate_func, data_type, data, ....) function that I could then call from a partial like validate_json = partial( validate, json.loads, "JSON")
or validate_uuid = partial(validate, uuid.UUID4, "UUID").
So what would be the difference between using closure and functools.partial? When would you prefer to use closure over partial?
you told me to buy Dutch bread after getting the free guide, but I see no link in video description for where to buy it, so I guess I will just keep living in my "bread fantasy world". Love your vids. They are always very informative. Thank you!
get_best_bread = partial(get_bread_from, country_code="de_DE") :) Great video! I was astounded seeing the closure part and then really impressed seing partial() in use. The flexibility and simplicity of providing values for named variables and using it on any function is amazing. My wife is gonna get jealous of the love I am developping for Python. Thanks for sharing this!
Man, i'm programming for 15years or more. Mainly c,cpp, python only 3y, but ive newer knows this closure and partial mechanisms! Great!
You have the cleanest code I’ve ever seen in my life!
Thank you for providing the github repo of the example. Nice!
You’re welcome!
Thank you for another great video Arjan! also, your anecdote reminded me of a similar pun "no matter how kind your kids are, German kids will always be kinder"
Is there any difference between using a partial function and a lambda function?
For some f(x,y), shouldn't lambda x: f(x, y0) be equivalent to partial(f, y=y0) ?
I like the last example with "partial functions". Still Composition, but very flexible. Thank You again for your teaching!
Thank you for the support, Alexa! Glad you're enjoying the content!
17:55 where did the partial grab the prices from however if it was not told..?
While I don't follow everything you do, I like to watch your videos because it exposes me to new ways of thinking as well as showing me the practical reality of what is typed to accomplish an objective. Very cool indeed. :)
I want more contents about functools. I love functional programming and i hope you will provide dedicated courses about it
Very informative! As someone who was previously unaware of partials, I have achieved similar functionality by setting constants and using them as default parameters (as you mentioned). This approach seems cleaner, though, so I will definitely give it a whirl in the future!
Excellent, as usual!
The partial function will be used tomorrow, replacing a (very cumbersome) workaround. Thanks👌
Glad to hear it was helpful, Johan!
12:35 But you could replace it with a docstring. Which could be dynamically generated, to incorporate the actual limit, and assigned to the ＿＿doc＿＿ property of the function object being returned.
I'm in love with your content. I think I'll purchase your course the SW designer mindset.
Keep it up my man!
Nothing really related to your video, but I wanted to thank you for your work. Not only it is very well explained but it really inspires me by debunking some dev "features" and mindset related to development. I don't really study your videos (sometimes still too advanced for me), but they show me every time the beauty of code and keeps me learning further. Furthermore, I tend to really enjoy your both light/humble and technical tone, I'm from marketing, and your channel is doing marketing right focusing on content value first. So again, thank you for your work :)
Just yesterday, it inspired me to have fun playing with inheritance and composition with PyMongo's source code (not touching at it, but making my way around it for some purpose). Trying to deep dive into code like you do, finding beautiful and fun features, far from just focused tech tutorials, more of a technical case study / learning experience. thx !
Do partials have any affect on time complexity of a function? Say you have a function where the part that is partially applied takes O(n^2), and the remaining part would be O(n). When calling the partially applied function, would this be O(n), because the O(n^2) part has already been handled or O(n^2) because it hasn't? This seems quite niche, but im currently working on graphical demonstration of time complexities and partials could be very helpful as I could only look at timing a specific part of a function. I watched this video a while back and remembered how useful I found it
Someone probably mentioned this already, but if you refactor the functions you're passing to partial so the arguments you're specifying therein are to the left of the one the new function still needs, you don't need to pass them by keyword.
I'm not sold on closures. Besides encapsulation what are they good for?
repeat and partial can come in really handy when it comes to multithreading and passing arguments to the executed function
What is the name of the theme you use in VS Code? I love the colour scheme of syntax highlighting in your videos.
Great video. I've started to use partial more and more in my functional code. I was using lambda for the same result, but partial is a lot better.
Interesting! What are the perks? In short perhaps...
@@martinleonhardt1541 Partial will work around mutable types and closure better. Lambda got me in a pickle when I tried using it in a list comprehension.
Once again, Arjan! Super helpful. Thanks!
I constantly write code that starts dead simple and expands in options progressively, ending up with very customizable functions. I’m gonna use partials to take care of this now!
Bedankt! Thank you, your video's are really helping me getting the most out of python and I really appreciate that as a self taught Pythonista :)
Great content, I work in the Netherlands as a Python Backend Dev and and the bread is okish haha!
You said 35_000_00 is thirty five thousand, but you also said that the underscores doesn't change anything about the integer value so it would be 3500000 so that wouldn't add 2 decimals randomly? So it should be 3_500_000 which is 3,5 million?
No, no, no - you definitely did not fall victim to the Dunning-Kruger effect (about bread), you just have a very good sense of humor.
This is also a great example of where you could also use a state machine :)
Love your videos man
Thank you!
How would you do dependency injection for functions?
For instance I have several functions where I want to use a database, for this I have a class or function which I want to re-use and mock for unit tests.
If I use classes I would inject this dependency in the constructor. In the composition root I would assemble my objects. If I would do the same on function level for instance with te partial() it sounds like a lot of configuration, and a bit unclear for me how to structure the code then.
The `key` argument in list.sort() only accepts functions with a single input. Partial came to my rescue when i wanted to do some funky sorting.
Thanks
Thank you Martin, glad you liked the video!
Thanks!
Thank you so much for the support! I really appreciate it!
@@ArjanCodes I am a developer from Taiwan( in Asia). I have been writing Scala code as a data engineer for the last three years. Recently I joined a new company and my team uses Python for everything. And I struggle to accustom myself to the new environment. Watching your videos about Python’s usage and how to write code in a more functional programming way makes me more confident and more productive. Thanks for the effort!
Why did he put the extra 2 zeros on 35_000_00 ? It evaluates in my interpreter as 3,500,000 when I'm sure he meant 35,000. Could someone please explain what I'm missing?
I’m using the integer type to represent a monetary amount, and then the common way to do it is to use cents as the base unit and not dollars.
Great video as always. Keep going!
French bread is of course better by far than any other bread, but I guess it wouldn’t be fair to include it in the competition as there wouldn’t be any controversy.
I am English and I'm still willing to back you up on the superiority of French bread!
@@jasonmcclatchie6877 thanks hehe
My mother used to make bread with recipes that undoubtedly came from the Netherlands (as my country was a Dutch colony and before that we have no bread) and I'll say they're great! 👍🏽
I rarely comment on video tutorials, unless it gave me that "aha" moment, and this was one of those, thanks ArjanCodes for educating us!
btw, have you profile which "way" is more faster, the OOP or FP style?
I have studied industrial engineering in Spain, with little curriculum in programming. Is it possible that I prefer functional programming due to my background in engineering? I.e. decompose the 'problem' in little steps via functions and then there is a main function that calls functions which in turn call lower level functions...and so on.
While I use functools in many occasions I always wondered what the (technical) difference is between the partial function and creating a lambda (or even a function) simply reducing the arguments? I mean, i could also do
buy_strategy = lambda prices: should_buy_average(prices, 4)
While "partial" usually increases readability I wondered if there are any differences in terms of performance.
I am also interested in this.
Doing that, you use closure, but in simpler form.
@@marsma18 kind of, yes, but not in the way he showed closures in the video as my approach IS a function with a valid signature actually calling another function while the approach in the video RETURNS a newly generated function with a valid signature. Moreover I was interested in the technical difference than the logical difference.
@@comedyclub333 The only real difference that I came across was while implementing a naive save system in my game engine (naive in the sense that I just dump whole objects, including attached callback functions and everything to file using pickle). Now callback functions cannot typically have arguments, but often require a variety of them, so I end up using functools.partial fairly often - because lambdas are anonymous and cannot be pickled, while partial functions can.
@@RitchieDiamond That's actually something really interesting. I just tried it and you are right! It seems that lambdas really are not savable into serialized data. Crazy!
I sure learn a lot from your videos. Thank you so much!
You’re welcome, Alfonso!
Very useful if you want to write to a file using print:
with open("output.txt", "w") as f:
fprint = partial(print, file=f)
fprint("Hello world!")
Same for printing to stderr
import sys
eprint = partial(print, file=sys.stderr)
...
eprint(f"Error occurred: {error.description}")
Even if I'm not yet at the level where I have to implement most of the things you share, I still go ahead to watch your videos anyway. You teach well.
That said, I thought I should point out something to you (and to those not familiar with this style). At minute 12:51, you add two additional zeroes, presumably to make the number 35,000. But that actually changes the number to 3,500,000. (And I confirmed this on my python shell.)
35_000_00 != 35,000
Instead,
35_000_00 == 3,500,000
35_000 == 35,000
3_4 == 34 (just to show those who might not know that you don't always have to end with zeroes)
In essence, it seems to me you think adding two zeroes after the last underscore means python will treat the variable as a float. It won't. It'll just add two more zeroes to the integer.
If you want a floating point number:
35_000.00
1_2.3
12_34_567_8009.027
are all different valid examples.
Money is best dealt with in integers because floats will have errors dividing in binary. It is 35000.00. When returning values, you just have to deal with the extra 00 as cents.
Thank you for the great tutorial and all the interesting Python content on your channel!
Glad you like them!
I'm German and I'm eating delicious bread as I'm watching this and I'm highly offended.
Like many good things. It's really convenient and right when your team knows each other about it.
Why did you decide to use dataclass for TradingBot?
Hi, Your videos are awesome... one question though.
How do you place that run button beside the split editor right icon in vs code, are you using any plugins/extensions?
Man you are the best! Could you make a video about design patterns of libraries like django or scrapy?
I'm a bit confused. in your dataclass you are defining buy_strategy as a TradingStrategyFunction which you defined as that Callable.. But what's the difference between that and just calling the definition directly from within the class? Without self.function_name() just function_name()?
The reason I'm using a function and not a method, is that by defining them as functions, you can set them when you create the object (this is what I do in the main function). If the buy/sell strategies were methods, you would need to create a subclass for each combination of buy/sell strategy you want, so it's less flexible.
Great video! Its begs the question of when a functional approach is better then a objected oriented one.
I have used both together.
Nice video as usual! Question on partial. Before knowing about it, I used to do:
buy_strategy = lambda prices: should_buy_average(prices, window_size=4)
vs
buy_strategy = partial(should_buy_average, window_size = 4)
I think the result is the same. Are these methods equivalent, but partial with less code?
There are a few differences I guess. For example with partial you can change the should_buy_average function (as long as it still takes the window_size) and need not change the partial statement, whereas you would need to change the lambda statement because it explicitly uses the parameters the should_buy_average function expects.
You can also change the parameters' order of the function and the partial application still works.
Nice! Thanks for another great tutorial.
What is the purpose of using a dataclass over a normal class for the TradingBot? Seems odd to me, so genuinely asking.
Love your videos Arjan!
Thanks Nick, glad you like them!
Hi, please make series of advance level python. Your way of explanation is very cool and the set of example you used is very cool.
Hi Arjan, Please continue the Design Pattern Series, You can take one by one different design patterns with practical code for python. There is not much good content out in youtube.
I'm interested in what keys you're pressing. How do you: 6:06 forward-delete an argument including the following comma and space up to the next argument? 15:31 delete a whole line at a time incl. LF? 15:49 delete to the end of the line? 17:59 forward-delete up to the end of the argument list?
Also British people know that our bread is the worst so we don't even pretend it's good.
Hi Arjan. Thank you for the instructions on passing extra params using closures. I bookmarked this video when I first watched it because I suspected that it would be useful, and today it was exactly what I needed to add functionality to the callback function passed to ftplib.retrbinary()
10:38 I like to use a name like “def_should_buy_avg”. The extra “def” on the front looks like the keyword for defining a function -- which is what the function does.
What is "recording" on every line of your terminal?
Functional patterns are awesome and one of my favourite ways to structure code, u see it a lot in scala or heaven forbid R (their class like defns are terribad)
But scalas map, filter, reduce, foldleft, currying or partial hell maybe i just used clojures and they got tail recursive.
I had all of my steps be like functional but isolated environments be classes.
I think both are cool
Hey @ArjanCodes, unrelated question: What's this sweater you're wearing? Thanks!
I've been using partial() for similar applications before. Well explained.
I have just discovered the channel. This is Golden
Thank you, glad you like the content!
I just discovered a disappointing downside of functools partial - partially applied functions don't typecheck!
I use mypy to check my type hints. I have a type hint for a "validate_multiple" function that expects a list of Callables with a certain validator signature. I tried to pass a partial to it and got this error:
Argument 3 to "validate_multiple" has incompatible type "List[object]"; expected "Sequence[Callable[[Dict[Any, Any], ExecutionContext], Dict[Any, Any]]]
Apparently there's an ugly workaround that involves manually casting the partial back to the right type, but that's not satisfying at all. :(
Good stuff, thanks for this great video!
Yeah, I'm liking your videos. I'm not a software engineering but I use python a lot to do data science.
I'm from the US, and I can vouch for the excellent NL accent! ...and I'll grant you the bread assertion
I totally agree about bread, I think the Polish one is the best ;) Anyway, another great tutorial, thanks!
Forget where I saw it on youtube, but someone commented that most classes are basically "collections of partially applied functions". In that scenario where I have similar methods, I would rather have user instantiate a class once than do the partial applications repetitively for the different functions / methods
Always learning something from your videos. Great content
Awesome! Thank you!
Nice! I’ve been writing Python for ages and didn’t realize partial was in there. Re:bread you need to try a good shokupan …
Have I ever said how much I like your videos?
It's nice that you can replace interfaces and their implementations with partial functions. However I'm still concerned with what happens in cases where we need to do some error validation and type checking, any ideas?
Does partial() mess up autocompletion?
I think we need an ArjanBakes channel where you show off Dutch bread
Hey. Speaking of design guide. I am signed up and get your notices at work but never got the guide. Is there a way to remove-request or access it ?
Hi Greg, send me an email (business@arjancodes.com) and I’ll sort it out for you.
Will do
Isn't it possible to use lambda functions instead of partial?