IRC logs for #openttd on OFTC at 2024-06-11
⏴ go to previous day
00:12:34 <audigex> andriydohniak: After 20 years of development I'd say that DOESN'T look nice
00:12:34 <audigex> Like sure, it's a clever bit of code... but as far as I'm concerned code should be something you can pick up and read and understand from reading it (assuming a baseline competence)
00:12:34 <audigex> You shouldn't have to either know what it does ahead of time, or "figure it out" excessively. To me that's a sign of bad code and I regularly reject PRs at work from my juniors for trying to be too fancy. If you can't hand it to a graduate developer who's been with us for 3 weeks and expect them to understand what it does (albeit perhaps not *why*), it's too fiddly
00:14:42 <audigex> Far too many developers put a weird focus on clever code that does nothing but prove how clever they are. Occasionally you have to do something fiddly looking for performance reasons, but 99% of the time you're better off being more verbose and writing understandable, maintainable code first and foremost
00:14:42 <audigex> Lines of code are cheap, maintenance hours are expensive
01:23:15 <bigyihsuan> andriydohniak: without looking at the replies
01:23:15 <bigyihsuan> going inside to out
01:23:15 <bigyihsuan> 0 if top <= bottom, otherwise
01:23:15 <bigyihsuan> ints from bottom to top+1 exclusive, then take evens, then sum together
01:23:15 <bigyihsuan> so the sum of even numbers within a given range, and 0 if the range is invalid
01:24:36 <bigyihsuan> audigex: my workplace uses go for our apis, and i can say to you that all go code looks the same
01:31:59 *** Wormnest has quit IRC (Quit: Leaving)
02:09:38 *** gnu_jj_ has quit IRC (Ping timeout: 480 seconds)
02:19:49 *** debdog has quit IRC (Ping timeout: 480 seconds)
03:04:17 *** gnu_jj_ has joined #openttd
03:07:28 *** gnu_jj has quit IRC (Ping timeout: 480 seconds)
03:31:54 <ahym> bigyihsuan: I thought it was a clever use of a ternary as well
03:32:31 <ahym> in terms of optimization i don't think you save any processing time doing it that way versus a conditional but i could be wrong
03:34:25 <bigyihsuan> ahym: all the ternary does here is being a "is the range valid" check, in go this'd be a bog standard if-return-else statement
03:35:40 <ahym> bigyihsuan: it condenses it slightly, there's other, arguably easier to read ways to write that in c++ that would still be performant as well
03:36:09 <ahym> I'd have to run a disassembly to know exactly how many comparisons it saves but I'm assuming zero
03:37:37 <ahym> If you wanted to save time this could be packed into a 128 register and probably be done quicker but it would be even less readable
03:42:00 <ahym> You could definitely use xxmintrin here to add more numbers within the range at once, it should reduce the instruction count at least by 4
05:28:11 *** keikoz has quit IRC (Ping timeout: 480 seconds)
06:06:43 <peter1139> Hmm, strange, Firefox keeps stalling watching videos on Youtube.
06:43:20 <silent_tempest> Do you have ad block?
07:11:59 *** SigHunter has joined #openttd
07:36:42 <peter1139> There is constinit in C++20, I guess.
07:51:44 <peter1139> Oof, group_gui.cpp:1175 and vehicle_gui.cpp:2241 are nasty.
08:08:38 *** SigHunter has quit IRC (Remote host closed the connection)
08:09:39 *** SigHunter has joined #openttd
09:56:31 *** Flygon has quit IRC (Quit: A toaster's basically a soldering iron designed to toast bread)
10:19:49 <peter1139> Well, still hungry :(
10:33:34 <peter1139> I ate them all yesterday.
10:54:44 * pickpacket had lunch with an old acquaintance
10:54:55 <pickpacket> That spelling looks wrong
11:15:44 <peter1139> Lots of discussion eh?
11:16:12 <pickpacket> not as much as the forum threads about it, though
11:18:00 <peter1139> I'm betting none of this gets resolved down to a defined feature list, let alone someone willing to implement it.
11:19:40 <peter1139> Feature/functionality.
11:19:49 <pickpacket> definitely not! But one can hope to gain something from it
11:57:25 <andriydohniak> How does fmt::println work? How do I print a simple int?
12:00:17 <andriydohniak> ok, I figured it out, its {:d}
12:08:54 <andriydohniak> Can I somehow create a separate window, that talks to it's "parrent" without using global statics?
12:09:43 <andriydohniak> or can I somehow store the window object in another window object and get access to it's memory & methods
12:18:27 <andriydohniak> FindWindowById, and then cast the pointer to the correct type?
12:18:52 <andriydohniak> I see that as the best option, because that is the only way I can be sure it's not closed
12:20:10 <andriydohniak> now I get why everybody uses globals
12:20:15 <andriydohniak> just less things to worry about
12:20:32 <peter1139> dynamic_cast<>(FindWindowById()) is used in many places already.
12:20:52 <peter1139> But also, some other windows use a static class for state.
12:21:19 <andriydohniak> that might be a good idea, but it also makes all state of that class shared
12:21:32 <andriydohniak> so I can't have different selections on different graphs for example
12:22:18 <andriydohniak> I am not sure if it will be ok to have all company selectors share the same state
12:23:07 <peter1139> If you can have different state, then you need some way of visually linking your company selector to the window it applies to.
12:23:53 <peter1139> If you have two company selectors open, which is which?
12:23:57 <andriydohniak> True, I was thinking most of them will be "embeded" like a widget inside the same window, so no issue there
12:24:02 <andriydohniak> peter1139: true
12:24:20 <peter1139> If it's embedded, then it's not a separate window and there's no state to pass around.
12:24:36 <andriydohniak> Mb I just make all of them embeded?
12:26:07 <andriydohniak> Can I make a window change size, for example on a button click, to show the selector?
12:26:16 <andriydohniak> so it doesn't always take up space?
12:26:31 <peter1139> Yes, the shade button is an example of that.
12:27:04 <andriydohniak> And can I add widgets dynamically to the window
12:27:24 <andriydohniak> or do I just make it always be there, and just set the relavent size to 0
12:28:11 <peter1139> Always there but hidden. But you can add them to the layout dynamically anyway so that the widget structure is shared.
12:28:32 <andriydohniak> any examples of this in the current UI?
12:28:33 <peter1139> Like how the exist company buttons are added.
12:28:44 <peter1139> Or the picker_gui widgets.
12:29:21 <peter1139> Bah, Doom maps that are just non-surprise trap-after-trap :(
12:29:39 <andriydohniak> peter1139: you are a true retro gamer :)
12:30:00 <andriydohniak> I also played through doom once, on medium difficulty, it was fun while it lasted
12:30:15 <andriydohniak> I also used an engine with mouse support, so a bit cheaty
12:30:42 <peter1139> Doom always supported mouse out of the box, so not really.
12:30:57 <peter1139> We just were used to playing it that way...
12:31:08 <andriydohniak> peter1139: but from what I remember stock mouse support is kinda crap, and you can't look up
12:31:24 <peter1139> Looking up and down is wrong, indeed.
12:31:45 <andriydohniak> I enjoyed being able to look up
12:38:13 <andriydohniak> I know why it was impossible to look up in the original, it used some weird 2d to 3d ish rendering scheme, but I guess modern doom engines just ignore that and render in full rd
12:54:37 <peter1139> It's a 3D game with a heavily optimised (it needed to run on a 386) renderer that can't look up & down.
13:53:43 <talltyler> Agh, my finger slipped and I clicked “Rebuild all” instead of “Build all” 🤦
14:04:01 <_glx_> Or similar combo, not sure as I'm not in front of a keyboard
14:12:25 <talltyler> Good trick, thanks. Also F7 for Visual Studio on Windows 🙂
14:13:22 <andriydohniak> make openttd :)
14:22:54 <kuhnovic> talltyler: Depends on which keyboard shortcut set you have configured, there's a few of them
14:23:45 <kuhnovic> And I can't recommend the Ctrl+F7 shortcut enough, which only compiles the current .cpp file. Saves lots of time.
14:25:22 <peter1139> I'm currently developing some back-end code in a front-end C# asp view.
14:25:43 <peter1139> Entirely so that I don't have to keep recompiling for every change.
14:26:07 <peter1139> (Views are compiled on demand when they changed, back end library code not so much...)
14:36:18 <LordAro> what a low effort piece of rubbish
14:36:37 <LordAro> "i don't have time/don't know how to" would've been fine
14:38:59 <peter1139> Say what you really mean :)
14:40:06 <_glx_> and I typed my reply while you were
14:41:55 <LordAro> peter1139: i already watered down the above :p
14:43:31 <LordAro> i'm surprised it even parses
14:43:46 <LordAro> i suppose cmake is fairly lax with undefined variables?
14:44:40 <peter1139> I relaxed it with "if", because I don't know if it actually works.
14:50:19 <peter1139> Urgh, incredibly verbose and repetitive XML :(
14:50:51 <LordAro> we solved that by bolting extensions to our xml parser
14:51:02 <LordAro> like includes & inheritance
14:51:11 <LordAro> it works as well as you might expect
14:52:16 <ahym> writing performant xml parsers is an art form
14:53:03 <LordAro> the fact that it's all xpath-based and continually recomputes/reparses stuff over and over doesn't much help
14:54:35 <peter1139> This is meant to be a standard document format, so it's flexible. And a pain.
14:55:53 <ahym> LordAro: This seems like a performance nightmare
14:56:24 <ahym> plaintext in general is too verbose
14:56:36 <LordAro> ahym: it's ok, we have a caching layer in front of it
14:56:37 <_glx_> <longTagName>a</longTagName>
15:28:06 <peter1139> #12777 is completely pointless anyway.
15:28:47 <peter1139> I also, uh, have some patches for this...
15:31:57 *** Wormnest has joined #openttd
15:43:26 <andriydohniak> How would I go about creating a widget, and setting it's size at 0, and changing that later. All my attempts were unsuccesfull
15:52:48 <_jgr_> NWID_SELECTION/NWidgetStacked may be something to look into for this
15:59:10 <andriydohniak> How and when is the size of widgets determined anyway?
16:00:43 <andriydohniak> and how do nested widgets work?
16:02:27 <andriydohniak> NWID_SELECTION is a pretty good choice, but then I will have to duplicate the widgets
16:04:38 <_jgr_> Probably the best places to look to see how sizing is done are in the various overrides of `NWidgetBase::SetupSmallestSize`, and in `Window::ReInit`
16:05:19 <_jgr_> And `Window::UpdateWidgetSize`
16:09:27 *** D-HUND is now known as debdog
16:44:41 *** HerzogDeXtEr has joined #openttd
17:49:28 <peter1139> Okay, `union Colour` is a bit scary with std::endian.
18:01:59 <peter1139> I've done it the verbose way.
18:05:28 <peter1139> May or may not compile :p
18:05:36 <peter1139> May slow compilation by 100x.
18:11:56 <peter1139> In my defence, the CI does build with clang like wot I iz using.
18:49:55 <truebrain> upgrade to C++23 when?
18:51:17 <Rubidium> for just std::byteswap we might be able to get away with it
18:51:24 <LordAro> i mean there is __builtin_bswap64
18:51:53 <exceptik> Isn't it gnu specific?
18:51:58 <LordAro> _byteswap_uint64 for MSVC
18:56:01 <peter1139> The MSVC built-ins are not compatible with constexpr.
18:56:21 <LordAro> well that's their loss
18:56:47 <truebrain> I just read you only need to fix 1 target
18:56:52 <truebrain> the others are ... "someone elses problem"
18:57:17 <LordAro> regardless, byteswap isn't particularly a complex function, it's not like the maintenance overhead is massive
18:58:00 <truebrain> I just now invision peter writing a 1000 line function to byte-swap, just to annoy you 😛
18:59:15 <peter1139> Convert it to json-packed XML with some base64 thrown in?
19:00:04 <peter1139> I remember that XML BACS STD18 file format...
19:05:44 <peter1139> (STD18 is a fixed-length text file format. The XML version of it was... a fixed-length text file format wrapped in some XML tag...)
19:11:56 <andriydohniak> Can somebody explain this c++ thing?
19:11:56 <andriydohniak> static WindowDesc _graph_legend_desc(
19:11:56 <andriydohniak> WDP_AUTO, "graph_legend", 0, 0,
19:11:56 <andriydohniak> WC_GRAPH_LEGEND, WC_NONE,
19:11:58 <andriydohniak> std::begin(_nested_graph_legend_widgets), std::end(_nested_graph_legend_widgets)
19:12:04 <andriydohniak> Is this a function? a constructor?
19:12:12 <andriydohniak> it's probably a constructor, right?
19:12:55 <LordAro> all of the window stuff is "a bit weird" as far as C++ goes
19:13:11 <LordAro> autoformatters hate it :D
19:13:40 <andriydohniak> And why can't I do a similar thing inside a struct, and I have to use {} instead of ()?
19:14:46 <LordAro> you mean in a struct definition?
19:14:53 <LordAro> that's not a constructor
19:15:07 <LordAro> the above is constructing an instance of WindowDesc
19:15:18 <LordAro> struct FOO { ... } is defining a struct
19:16:03 <LordAro> so your question doesn't really make sense
19:16:06 <andriydohniak> and what does this do then?
19:16:06 <andriydohniak> WindowDesc _graph_legend_desc(
19:16:06 <andriydohniak> WDP_AUTO, "graph_legend", 0, 0,
19:16:08 <andriydohniak> WC_GRAPH_LEGEND, WC_NONE,
19:16:10 <andriydohniak> std::begin(_nested_graph_legend_widgets), std::end(_nested_graph_legend_widgets)
19:16:18 <andriydohniak> I need to change something
19:16:37 <andriydohniak> WindowDesc _graph_legend_desc{
19:16:37 <andriydohniak> WDP_AUTO, "graph_legend", 0, 0,
19:16:37 <andriydohniak> WC_GRAPH_LEGEND, WC_NONE,
19:16:38 <andriydohniak> std::begin(_nested_graph_legend_widgets), std::end(_nested_graph_legend_widgets)
19:16:44 <andriydohniak> the first desn't work
19:17:21 <LordAro> aha, you've run into C++ uniform initialisation rules
19:18:54 <LordAro> but, what would work for your first one is `struct foo { WindowDesc my_desc = WindowDesc(WDP_AUTO, ...); };`
19:19:10 <LordAro> initialising members "inline" in struct definitions is still kinda new
19:20:03 <LordAro> previously you would've had to define it in the struct's constructor
19:20:51 <andriydohniak> I will do that :)
19:21:10 <andriydohniak> I tried with the new keyword, but the copy constructor is explicitly deleted
19:21:17 <andriydohniak> for some structs*
19:22:50 <LordAro> ( with "non-static data member with initializer" )
19:23:25 <andriydohniak> and what about static initializers?
19:23:40 <andriydohniak> Can I initialize const static members inline?
19:24:01 <LordAro> that's in the above too :p
19:33:32 *** gelignite has joined #openttd
19:37:44 <andriydohniak> How can I reuse an array of NWidgetParts in another array of NWidgetParts
19:38:59 <LordAro> the same way you'd insert an array into any other array, i imagine
19:39:25 <LordAro> (i.e. probably a vector)
19:40:38 <andriydohniak> I can't use some sort of spread operator?
19:41:12 <LordAro> well, maybe if you try hard enough with some template magic
19:41:22 <andriydohniak> No, I don't want that :)
19:42:00 <andriydohniak> NO, NO, NO GOD PLEASE! NO!
19:42:09 <peter1139> andriydohniak, look at how picker_gui.cpp does it.
19:42:20 <andriydohniak> peter1139: what line?
19:42:29 <peter1139> Somewhere in the bottom.
19:42:56 <peter1139> There are Make* functions that create the widgets, and then other widget trees call those functions.
19:44:04 <andriydohniak> peter1139: Thank you!
19:44:39 <andriydohniak> there is a MakeNWidget, that takes in a couple of iterators, and returns a NWidgetBase
19:44:46 <andriydohniak> that can be used in another array
19:46:02 <peter1139> Yes, it inserts the widget tree at run-time.
19:46:56 <peter1139> I've also got a patch that simplifies the begin/end stuff a bit.
19:47:43 <andriydohniak> How long do you think untill upstream?
19:52:23 <LordAro> (a recurring meme is that peter *always* has a patch for "that")
19:52:30 <LordAro> could be a lot sooner, mind
19:55:17 <peter1139> The MakeWidget and MakeWidgetTree changes are slightly odd. We need to update the beginning of the span, but you can't do that with spans. So I turned the span into a pair of iterators and use those to make a new span...
19:55:34 <peter1139> subspan() is a thing but now really suitable here.
19:56:10 <peter1139> And keeping those functions as `const NWidgetPart *` and passing span.begin()/end() doesn't work, because the type different.
19:56:18 <peter1139> `&*span.begin()` does work, but ugly.
19:57:06 <andriydohniak> I don't get what half of theese words are, so I wish you good luck, and if it touches any code I write, I will gladly learn and update my code then
19:58:32 <peter1139> I basically replace those two pointers, begin and end, with a span.
19:58:44 <peter1139> Which is more C++-like, as pointers are bad.
20:06:00 <andriydohniak> Why can't it just accept an array? ahh, because c++ array is a pointer, and we can't know if it was allocated
20:06:48 <peter1139> Because an array only tells you the start of it.
20:07:04 <peter1139> A span is both the begin and end.
20:08:34 <andriydohniak> forgot about that too :)
20:08:47 <andriydohniak> but std::array or std::vector would work, right?
20:09:09 <andriydohniak> but spans are immutable, so technically more versatile
20:09:23 <andriydohniak> or mb they aren't, idk c++ is hard
20:10:09 <andriydohniak> Anyway, what order do the functions get called and the UI gets created when a new window is created
20:12:06 <andriydohniak> And can I still use functions like GetScrollbar for the widget that is nested via NWidgetFunction
20:36:45 *** gelignite has quit IRC (Quit: Stay safe!)
21:09:14 *** nielsm has quit IRC (Ping timeout: 480 seconds)
21:22:19 *** keikoz has quit IRC (Ping timeout: 480 seconds)
21:24:07 <peter1139> OnInitialPosition() is pretty clearly broken :(
21:35:41 <peter1138> Optimised gameplay?
21:42:32 *** ChanServ sets mode: +v tokai
21:49:17 *** tokai|noir has quit IRC (Ping timeout: 480 seconds)
22:13:12 *** Wolf01 has quit IRC (Quit: Once again the world is quick to bury me.)
22:18:01 *** HerzogDeXtEr has quit IRC (Read error: Connection reset by peer)
22:19:51 *** HerzogDeXtEr has joined #openttd
22:23:32 <_glx_> andriydohniak: std::array can be annoying because compile time fixed size determination
23:29:45 <silent_tempest> I guess it's not possible for me to request review from others on my PR's?
23:30:06 <silent_tempest> Like requesting a review from a speciffic individual.
23:52:19 <silent_tempest> This^^ is tiny if someone is looking a quick review.
continue to next day ⏵
