on Lua
This commit is contained in:
parent
ebb4543f8d
commit
b6e803cfee
11 changed files with 162 additions and 30 deletions
60
content/programming/languages/cxx.tree
Normal file
60
content/programming/languages/cxx.tree
Normal file
|
@ -0,0 +1,60 @@
|
|||
% id = "01H9R1KJESR2F420HE67HW4AVR"
|
||||
- design lessons from the best programming language of all time that everyone loves (not really)
|
||||
|
||||
% id = "programming/cxx/access-modifiers-as-labels"
|
||||
+ access modifiers as labels (`private:`, `protected:`, and `public:`)
|
||||
|
||||
% id = "01H9R1KJES39Z6RBCKY4E71PYD"
|
||||
- although Java and C#'s approach to symbol privacy may be verbose, it has one great advantage: it is stateless.
|
||||
|
||||
% id = "01H9R1KJES17626QXYEGM7XBC7"
|
||||
- the way they're implemented in C++, it's essentially a bit more parsing state you have to keep track of
|
||||
|
||||
% id = "01H9R1KJESG4K8T1K1G36T7HBP"
|
||||
- and you know what other parsing state you have to keep track of in C++? - that's right, the preprocessor.\
|
||||
access modifiers, like all tokens, are affected by the preprocessor, and you have to take that into account.
|
||||
|
||||
% id = "01H9R1KJESJ0G0VQAW994ZHR0S"
|
||||
- take the following example:
|
||||
```cpp
|
||||
class ComfyZone
|
||||
{
|
||||
std::vector<SoftBed> _soft_beds;
|
||||
|
||||
#if ENABLE_HUGS
|
||||
|
||||
public:
|
||||
void hug(Person& person);
|
||||
|
||||
#endif
|
||||
|
||||
int _remaining_hugs = 10;
|
||||
};
|
||||
```
|
||||
|
||||
% id = "01H9R1KJESDDX4089WVHVV8N3H"
|
||||
+ although quite contrived, it illustrates the problem pretty well
|
||||
|
||||
% id = "01H9R1KJESD2KED5TAFBY426A6"
|
||||
- (before you ask, `_remaining_hugs` needs to be always present because it has to be (de)serialized no matter if hugging functionality is compiled in. otherwise we'd get data loss.)
|
||||
|
||||
% id = "01H9R1KJESES27VKVW4A0ZVM11"
|
||||
- we intended for `_remaining_hugs` to be private, but if hugs are enabled, it becomes public.
|
||||
|
||||
% id = "01H9R1KJESTKW90R788SSPMNC6"
|
||||
- this can be _very_ hard to spot if you have a big class with lots of declarations inside.
|
||||
|
||||
% id = "01H9R1KJESCJ3VC8ATPYFDCPSP"
|
||||
- this can be worked around by banning access modifiers from appearing in `#ifdef`s, but you have to *realize* that this might happen
|
||||
|
||||
% id = "01H9R1KJES4ZYHVADDF80WAXH6"
|
||||
- and I've seen instances of this exact thing occurring in the Unreal Engine codebase, which is *full* of long lists of declarations (made even longer by the prevalence of `UPROPERTY()` specifiers)
|
||||
|
||||
% id = "01H9R1KJES182MCV2V0A4VHKKX"
|
||||
- even if we didn't have the preprocessor, that access modifier is state _you_ have to keep track of
|
||||
|
||||
% id = "01H9R1KJESH7PWNKCKW3H0WJHW"
|
||||
- I very often find myself needing to scroll upward after <kbd>Ctrl</kbd>-clicking on a field or function declaration, just to find out if I can use it
|
||||
|
||||
% id = "01H9R1KJESFE6F1D4J5PA5Q381"
|
||||
- (thankfully IDEs are helpful here and Rider shows you a symbol's visibility in the tooltip on hover, but I don't have Rider on code reviews)
|
74
content/programming/languages/lua.tree
Normal file
74
content/programming/languages/lua.tree
Normal file
|
@ -0,0 +1,74 @@
|
|||
- TODO: this page could really use an interactive Lua interpreter. can we have that?
|
||||
|
||||
- Lua is a really cool language! did you know that?
|
||||
|
||||
+ lots of people complain about it being really weird for various reasons, but these are generally superficial
|
||||
|
||||
- usually it's cosmetic stuff, so these aren't any arguments of technical merit, but...
|
||||
|
||||
- stuff like indexing from 1 instead of 0, which is _just a design choice_ and does not impact your programming that much
|
||||
|
||||
- in fact, one could argue that regular programmers are weird for counting from zero :thinking:
|
||||
|
||||
- or using `~=` instead of `!=`, which is _just a syntax choice_, and you only have to get used to it once
|
||||
|
||||
- or using `do`..`end` style blocks instead of `{`..`}`, which again is _just a syntax choice_ and does not impact programming that much
|
||||
|
||||
- it's a tad bit more line noise, but not that terrible. I [did design a language using `do`..`end` blocks][def:mica/repo] and it really doesn't look that bad
|
||||
|
||||
- but I think Lua is a pretty damn genius programming language.
|
||||
|
||||
- the use of tables as The One Data Structure for Literally Everything strikes me as a 200 IQ choice I could never come up with myself
|
||||
|
||||
- partly because it's so fucking bold I can literally not imagine myself designing a language with a strong distinction between hash tables and arrays, and even tuples and records!
|
||||
but the authors of had the restraint to just have One. and that One is **tables.**
|
||||
|
||||
- tables are extremely powerful in what they can do, because they're more than just a way of structuring data - they also allow for interfacing with the language *syntax* through operator overloading
|
||||
|
||||
+ in fact object oriented programming in Lua is typically done by overloading the `[]` indexing operator.
|
||||
|
||||
- the way it works is that `a.b` is just syntax sugar for `a["b"]`, which means you overload `[]` to _fall back to another table_ - and that way you can achieve [prototype-based inheritance](https://en.wikipedia.org/wiki/Prototype-based_programming)!
|
||||
|
||||
```lua
|
||||
local fallback = { b = 2 }
|
||||
local base = { a = 1 }
|
||||
|
||||
-- The __index field can be both a function *and* a table.
|
||||
-- { __index = the_table } is a shorthand for { __index = function (t, k) return the_table[k] end }
|
||||
setmetatable(base, { __index = fallback })
|
||||
assert(base.b == 2)
|
||||
```
|
||||
|
||||
- TODO: even more on restraint (standard library)
|
||||
|
||||
- TODO: restrained sugar: function calls `require "abc"`, `do_stuff { with_a_table = 1 }`
|
||||
|
||||
- I really wish Lua had at least *a* form of static typing though, since knowing about errors you make early is _really_ helpful during development.
|
||||
|
||||
- it regularly happened to me that a type error I made only occured at *some point* later during runtime; and then you have to track down a reproduction case and make a fix at the source. not fun.
|
||||
|
||||
- and it's really a bummer that Lua is not that strict!
|
||||
|
||||
- global variables by default are a pretty bad design choice in my opinion. having any form of uncontrolled globals hurts local reasoning and makes it harder to tell whatever your code is going to do.
|
||||
|
||||
- but fortunately it is possible to freeze your global variables by setting a metatable on `_G`, which is a table that represents the global scope.
|
||||
|
||||
- TODO: example
|
||||
|
||||
- there are also some bits of syntax that haven't aged very well.
|
||||
|
||||
- as much as people complain about cosmetics [TODO: link], I think there's a particular design choice that has aged very poorly in the face of modern, functional programming - function literals.
|
||||
|
||||
these tend to be quite verbose in Lua which hurts readability in functional code:
|
||||
|
||||
```lua
|
||||
local u = map(t, function (v) return v + 2 end)
|
||||
```
|
||||
|
||||
compare that to JavaScript's arrow functions `=>`, which I think are a prime example of good syntax sugar that encourages more function-oriented programming:
|
||||
|
||||
```javascript
|
||||
let u = t.map(v => v + 2)
|
||||
```
|
||||
|
||||
- the lack of a pipelining operator `|>` is also an annoyance, albeit most modern imperative languages don't have it either.
|
Loading…
Add table
Add a link
Reference in a new issue