Log for #openttdcoop.devzone on 13th April 2010:
Times are UTC Toggle Colours
00:04:04  <PeterT> Ammler: I have a potential script fro you
00:31:34  *** welshdragoon has quit IRC
00:33:45  *** welshdragon has joined #openttdcoop.devzone
01:57:04  *** welshdragon is now known as welshdragon`
05:38:52  *** planetmaker has joined #openttdcoop.devzone
06:54:24  *** SmatZ has joined #openttdcoop.devzone
06:55:55  *** pm has joined #openttdcoop.devzone
06:56:25  *** tneo has joined #openttdcoop.devzone
07:04:13  *** ODM has joined #openttdcoop.devzone
07:05:01  *** SmatZ has quit IRC
07:05:01  *** tneo has quit IRC
07:05:01  *** pm has quit IRC
07:33:55  *** pm has joined #openttdcoop.devzone
07:35:25  *** tneo has joined #openttdcoop.devzone
07:35:55  *** SmatZ has joined #openttdcoop.devzone
09:14:57  *** planetmaker is now known as Guest2349
09:14:57  *** pm is now known as planetmaker
09:54:51  *** Doorslammer has joined #openttdcoop.devzone
10:20:30  *** ODM has quit IRC
10:25:22  <Doorslammer> Anyone have a location for "New Bridges v0.41"?
10:25:36  <Doorslammer> Searched the forum, it doesn't readily come up
10:27:50  <Doorslammer> Oh, its OK
10:33:01  <Rubidium> grfcrawler?
10:42:49  <Ammler> @wiki GRF_Table
10:42:59  <Webster> GRF Table 7.3 - #openttdcoop Wiki -
11:04:59  *** KenjiE20 has joined #openttdcoop.devzone
11:50:24  *** welshdragon` is now known as welshdragon
13:33:39  *** ODM has joined #openttdcoop.devzone
14:55:22  *** Seberoth has joined #openttdcoop.devzone
15:22:25  *** frosch123 has joined #openttdcoop.devzone
15:25:21  <Hirundo> Yexo: Shouldn't Variable inherit from Expr?
15:25:46  <Yexo> theoretically yes, but it doesn't matter much
15:27:44  <Hirundo> I know :) Expr doesn't do anything, basically
15:29:05  <Hirundo> Currently I'm working on evaluating expressions compile-time, is that OK?
15:30:54  <Yexo> sure
15:31:22  <Yexo> do you have a plan how to deal with signed/unsigned numbers?
15:33:10  <Hirundo> Good point, I'll test how pythons native functions in do that
15:33:40  <Yexo> I think in python every number is signed, but don't quote me on that
15:34:23  *** Seberoth has quit IRC
15:34:59  <Hirundo> hmm.. unary minus is not (yet) supported?
15:35:06  <Yexo> nope
15:35:39  <Yexo> and you can't just add support for it as it'll break the parser
15:37:18  <Hirundo> at least, it seems print_xxx fails on negatives
15:37:45  <Yexo> that is the easy part
15:38:08  <Yexo> just add if value < 0: value += max_value before the assert line
15:38:43  *** Seberoth has joined #openttdcoop.devzone
16:20:54  <Hirundo> Python seems to handle both > UINT32_MAX and < 0 without problems
16:40:50  *** Pe1erT has joined #openttdcoop.devzone
16:45:05  *** Pe1erT has quit IRC
16:59:10  <Hirundo> I'll commit the print_xxx part
17:06:02  <Yexo> ok
17:15:44  <Hirundo> For the rest, we need to fundamentally decide how to handle signed/unsignedness
17:17:55  <frosch123> is there a problem with doing everything signed and 32bitty?
17:18:39  <Yexo> the problem is how to write nfo from that
17:20:11  <frosch123> sounds like i miss something obvious :s
17:20:38  <Rubidium> frosch123: m4nfo meets python
17:21:57  <Yexo> every time I'm thinking about implementing it I think of several problems, but I can't remember those now :(
17:22:42  <frosch123> you might have trouble with sign-extending variables shorter than 32bit
17:23:55  <Hirundo> If you accept everything as signed, stuff like and-masks may give troubles since you need to go up to 0xFFFFFFFF for those
17:24:15  <Webster> Latest update from devactivity: NFO Meta Language - Revision 51: Codechange: Let print_xxx functions accept negative numbers with... <> || Redmine - Revision 3511: Fixes a test failure with SQLite3 due to transaction fixtures (#5290). <>
17:25:09  <Yexo> that shouldn't be a problem
17:25:22  <Yexo> only problem I can think of is variables that use 32bits
17:27:40  <Hirundo> Alternative is to provide duplicate operators where necessary (/, %, min/max), but it's not exactly nice
17:28:20  <Yexo> let's go for everything signed 32 bits then, we can provide the duplicate operators later if necesary
17:46:54  <Hirundo> in signed 32 bit arithmetic, 0xFFFFFFFF is interpreted as -1, or does it give an error?
17:48:52  <Rubidium> -1, assuming 2 complement
18:15:38  *** welshdragon has quit IRC
18:29:02  *** Seberoth has quit IRC
18:44:07  *** Seberoth has joined #openttdcoop.devzone
19:03:03  *** Hirundo has quit IRC
19:05:18  *** Hirundo has joined #openttdcoop.devzone
19:25:28  <Hirundo> Yexo: I narrowed down the parser errors wrt unary minus
19:26:04  <Hirundo> 'switch_range : expression COLON RETURN expression' <- more than one of those places two expressions next to each other, without anything in between
19:26:21  <Yexo> yes, indeed
19:27:16  <Hirundo> I guess this is the reason C requires semicolons everywhere :)
19:27:24  <Yexo> yeah :)
19:28:12  <Hirundo> hmm... how does python do this *tests*
19:28:36  <Yexo> by requiring a newline everywhere?
19:29:19  <Hirundo> Indeed
19:32:51  * Hirundo ponders semicolons
19:35:43  <Yexo> feel free to add that
19:37:02  <Hirundo> I'm studying the parser what consequences that will have everywhere
19:37:25  <Hirundo> How does the cargo table work, do you specify cargo labels like 'PASS' as IDs?
19:41:41  <Yexo> yes
19:42:06  <Hirundo> An then, you reference to cargos by the index in that table
19:43:22  <Rubidium> *cough*action6*cough* :)
19:43:26  <Yexo> for now yes, but later by name
19:44:07  <Yexo> what's with action6?
19:45:00  <Rubidium> that's what "they'll" be whining about that is missing and such
19:45:16  <Rubidium> not that many sane people use it
19:45:32  <Yexo> it's supported in several places, although indirectly
19:46:16  <Yexo> without action6 you can't do things like param[param[10]]
19:46:55  <Rubidium> oh :) using action6 to make some stuff work; nice!
20:27:55  <DJNekkid> what is the usage of action6 btw?
20:33:34  <Hirundo> Its usage in nfo or its implementation in nml?
20:38:11  <Hirundo> Yexo: semicolons coming in..
20:47:26  <Hirundo> Yexo: Shall I implement proper functionality for unary operators, or stick with -x => (0 - x) for now?
20:47:47  <Yexo> it's easy to implement it properly
20:48:02  <Yexo> just follow the example from the ply docs
20:48:35  <Yexo> so definitely implement properly
20:48:43  <Hirundo> The parser is not the problem, I'm talking about he ast side of things
20:49:23  <Hirundo> t[0] = BinOp(code_to_op[t[1]], ConstantNumeric(0), t[2]) or t[0] = UnaryOp(code_to_op[t[1]], t[2])
20:50:27  <Yexo> BinOp is probably easiest for now
20:51:06  <Hirundo> OK, then I'm done :)
20:51:14  <Yexo> great :)
20:52:53  <Hirundo> pushed
20:55:59  <Yexo> for varaction2 we need to get a list of all supported variables and their size anyway
20:56:24  <Yexo> should we support the var numbers or should we give each var (or part of var in some cases) a unique name?
20:57:27  <Yexo> or (like currently for properties) both?
20:58:56  <Webster> Latest update from devactivity: NFO Meta Language - Revision 52: Change: Semicolons are now required at the end of declarations a... <>
20:59:24  <Hirundo> My suggestion would be like this *searches example*
21:00:55  <Hirundo> player_type => ((var[43] & 0x00FF0000) >> 16)
21:02:05  <Yexo> I don't fully understand it, what are you going to write in nml? just "player_type"?
21:02:16  <Yexo> or do you need to include that definition too?
21:03:11  <Hirundo> just player_type, and we abstract away the rest
21:03:31  <Yexo> ok :)
21:03:47  <Yexo> so the current VAR suport can be removed from the parser then
21:05:06  <Hirundo> We could leave it in for advanced users, who could use it to boost performance
21:05:25  <Hirundo> by e.g. not accessing the same variable twice, but caching it
21:05:51  <Yexo> not sure how relevant that would be
21:06:43  <Hirundo> Anyways, setting properties by number is possible as well
21:07:50  <Yexo> yes, but only for those properties where a name is defined
21:08:27  <Yexo> train weight for example is 2 byte-sized properties in nfo, I've already abstracted them in a single property
21:08:41  <Yexo> but that means you can't set the byte-sized properties by number anymore
21:09:08  <Hirundo> Because you need to know the size, action0 can only use known properties, but for varact2 there is no such restriction
21:09:48  <Yexo> with the exception that in variables where not all bits are defined the other bits are undefined
21:09:52  <Yexo> so should be masked away
21:09:54  *** Seberoth has quit IRC
21:10:13  <Yexo> you can't guarantee that for unkown variables
21:10:44  <Hirundo> That is the user's responsibility
21:11:04  <Yexo> ok
21:12:11  <Hirundo> Playing a bit of a devil's advocate here... if var[x] becomes a burden, just trash it :)
21:12:33  <Yexo> it's already implemented, so doesn't matter much
21:13:15  <Hirundo> And now for something completely different.... I came across this piece of python:
21:13:19  <Hirundo> w = int( ( w & 0x7fffFFFF ) | -( w & 0x80000000 ) )
21:13:45  <Hirundo> It truncates to 32 bit signed, apparently that's not trivial in python
21:14:02  <Webster> Latest update from devactivity: NFO Meta Language - Revision 53: Feature: Support the unary minus operator. <>
21:16:38  <frosch123> isn't the "|" too much?
21:18:06  <Rubidium> isn't the -(w & 0x80000000) == 0 ?
21:18:59  <Yexo> Rubidium: no, integers are not limited to 32bits
21:19:16  <Rubidium> are you sure?
21:19:18  <Yexo> yes
21:19:20  <Hirundo> Python does everything in unlimited-size integers if it becomes too big for 32 bit
21:19:31  <Rubidium> ah, okay
21:20:26  <Hirundo> So in order to do compile-time arithmetic, a proper simulation of a 32 bit is needed
21:28:33  <Yexo> ID = expression change is now in without problems
21:29:58  *** frosch123 has quit IRC
21:30:37  <Yexo> the list with action0 properties is part of actions/, but that's not really the best location I think
21:31:22  <Yexo> would moving the property list to actions/ be ok?
21:32:00  <Yexo> and then add action2/ for a list with valid action2 vars?
21:36:39  <Hirundo> go ahead
21:38:23  <Hirundo> In the meantime, I got the compile-time evaluation done, as long as both operands and result are truncated to 32bit everything works like a charm
21:38:56  <Yexo> nice :)
21:39:21  <Yexo> not pushed yet?
21:39:45  <Hirundo> Not yet
21:43:19  <Yexo> are you going to push it or can I commit the property move list first?
21:44:24  <Hirundo> It's done now, I had to rebase to a remote change I had missed
21:44:27  <Webster> Latest update from devactivity: NFO Meta Language - Revision 54: Change: an ID is now an expression <>
21:47:40  <Yexo> Hirundo: aren't there a lot more places whre reduce_expr could be used?
21:47:46  <Hirundo> Yes
21:50:13  <Hirundo> Real-life /me wants some sleep, and I will worry about that tomorrow :)
21:50:34  <Yexo> gn Hirundo
21:50:42  <Hirundo> goodnight
21:59:33  <Webster> Latest update from devactivity: NFO Meta Language - Revision 56: Change: move the action0 property lists to a seperate file <> || NFO Meta Language - Revision 55: Feature: Compile-time evaluation of assignment expressions when ... <>
22:15:42  *** Guest2349 is now known as pm
22:25:01  *** pm has quit IRC
23:00:28  *** ODM has quit IRC
23:09:08  *** KenjiE20 has quit IRC
23:23:28  *** welshdragon has joined #openttdcoop.devzone
23:46:21  <Webster> Latest update from devactivity: NFO Meta Language - Revision 57: Feature: min/max can take more then 2 arguments <>

Powered by YARRSTE version: svn-trunk