Xudo questions about WML.

[Xudo]

When I check stderr.txt, I face strange error.

Spoiler:

20091207 20:55:08 error engine: could not de-serialize unit: 'load_game_failed: Attempt to de-serialize a unit with an empty 'type' field:
__value = 0
[__array]
advances_to =
ai_special =
alignment = lawful
alpha = 256
attacks_left = 1
cost = 44
description =
die_sound = human-die-1.ogg,human-die-2.ogg,human-die-3.ogg
experience = 0
facing = se
flag_rgb = magenta
flying = no
gender = male
goto_x = 0
goto_y = 0
halo =
hitpoints = 80
id = Guardsmen
image = units/human-loyalists/halberdier.png
language_name = Halberdier
level = 3
max_attacks = 1
max_experience = 150
max_hitpoints = 80
max_moves = 5
movement = 5
movement_type = smallfoot
moves = 5
name = Guardsmen
overlays =
profile = portraits/humans/halberdier.png
race = human
random_gender = no
random_traits = no
resting = no
role =
side = 2
type = Halberdier
undead_variation =
underlying_id = 8
unrenamable = no
upkeep = full
usage = fighter
variation =
x = 20
y = 18
zoc = yes
[defense]
castle = 40
cave = 60
flat = 60
forest = 50
frozen = 80
fungus = 50
hills = 50
mountains = 40
reef = 70
sand = 70
shallow_water = 80
swamp_water = 80
village = 40
[/defense]
[movement_costs]
castle = 1
cave = 2
flat = 1
forest = 2
frozen = 3
fungus = 2
hills = 2
mountains = 3
reef = 2
sand = 2
shallow_water = 3
swamp_water = 3
village = 1
[/movement_costs]
[resistance]
arcane = 80
blade = 100
cold = 100
fire = 100
impact = 100
pierce = 60
[/resistance]
[advancement]
description = Max HP bonus +3, Max XP +20%
id = amla_default
image = misc/icon-amla-tough.png
max_times = 100
strict_amla = yes
[effect]
apply_to = hitpoints
heal_full = yes
increase_total = 3
[/effect]
[effect]
apply_to = max_experience
increase = 20%
[/effect]
[effect]
apply_to = status
remove = poisoned
[/effect]
[effect]
apply_to = status
remove = slowed
[/effect]
[/advancement]
[portrait]
image = portraits/humans/transparent/halberdier.png
mirror = false
side = left
size = 400
[/portrait]
[portrait]
image = portraits/humans/transparent/halberdier.png
mirror = true
side = right
size = 400
[/portrait]
[variables]
[/variables]
[status]
hidden = yes
[/status]
[attack]
damage = 16
description = halberd
icon = attacks/halberd.png
name = halberd
number = 3
range = melee
type = pierce
[specials]
[firststrike]
description = First Strike:
This unit always strikes first with this attack, even if they are defending.
id = firststrike
name = firststrike
[/firststrike]
[/specials]
[/attack]
[attack]
damage = 20
description = halberd
icon = attacks/halberd.png
name = halberd
number = 2
range = melee
type = blade
[/attack]
[modifications]
[trait]
female_name = resilient
id = resilient
male_name = resilient
name = resilient
[effect]
apply_to = hitpoints
increase_total = 4
[/effect]
[effect]
apply_to = hitpoints
increase_total = 1
times = per level
[/effect]
[/trait]
[trait]
female_name = strong
id = strong
male_name = strong
name = strong
[effect]
apply_to = attack
increase_damage = 1
range = melee
[/effect]
[effect]
apply_to = hitpoints
increase_total = 1
[/effect]
[/trait]
[/modifications]
[/__array]

All pieces of code, what can be linked to this problem:

Spoiler:

Code: Select all

[event]
name=prestart
	[unit]
		type="Halberdier"
		id=Guardsmen
		name= _ "Guardsmen"
		side=2
		x,y=20,18
	[/unit]
	[store_unit]
		variable=floor[6][1]
		[filter]
			type=Halberdier
			x,y=20,18
			side=2
		[/filter]
		kill=no
	[/store_unit]
[/event]

Code: Select all

[event]
	name=moveto
	first_time_only=no
	[if]
		{VAR_EQ unit.x 21}
		{VAR_EQ unit.y 25}
		[then]
			[unstore_unit]
				variable=floor[6][1]
				x,y=20,18
			[/unstore_unit]
		[/then]
	[/if]
[/event]

What can be a reason for this error?

[SkyOne]

To me, that error message usually appears when something is wrong on the unit type(s) in the [side] tag(s), but I am not really sure.

[zookeeper]

Unsurprisingly, it's this bit:

Code: Select all

variable=floor[6][1]

That's simply not valid syntax. There are no multi-dimensional arrays in WML. If you need a 2-dimensional array, then you need to use something like variable=floor[6].foo[1].

However, the last time I tried, it was impossible to store a unit into a sub-array in the first place. That is, variable=foo[1] simply won't work. I'm sure you can work around that by storing it into a regular variable/array first (variable=foo) and then using [set_variables] to copy it into whatever sub-array you want to place it in.

[Gambit]

It is also possible, though a bit complicated, to turn units into strings and rebuild them.

[Xudo]

Today I found, what BfW don' want to store units to cells of array.

Code: Select all

[store_unit]
variable=foo[1]
...
[/store_unit]

Code: Select all

[store_unit]
variable=garbage
...
[/store_unit]
{VARIABLE foo[1] $garbage}

wrong syntax.(cause error)

Code: Select all

[store_unit]
variable=foo
...
[/store_unit]

right syntax.
It's good, what know what units will be stored and can create individual variables.

Thanks to all, who answered my post.

[Sapient]

However, the last time I tried, it was impossible to store a unit into a sub-array in the first place. That is, variable=foo[1] simply won't work. I'm sure you can work around that by storing it into a regular variable/array first (variable=foo) and then using [set_variables] to copy it into whatever sub-array you want to place it in.

It definitely should work. If not, please file a bug report.
FYI, some lesser used features of WML may have been broken by the extensive lua-related changes in 1.7 branch.
The sooner these bugs get reported, the better.
(So we can have a good stable 1.

[silene]

It definitely should work. If not, please file a bug report.
FYI, some lesser used features of WML may have been broken by the extensive lua-related changes in 1.7 branch.

Come on, stop putting the blame of every WML failure on Lua...

How can you expect [store_unit] to store several units (!) into the single variable foo[1]? As zookeeper already explained, there are no multidimensional arrays in WML. So this can't work and this doesn't have anything to do Lua (and this is actually caused by your code written in 2007, e.g. commit 19367).

[Sapient]

It definitely should work. If not, please file a bug report.
FYI, some lesser used features of WML may have been broken by the extensive lua-related changes in 1.7 branch.

Come on, stop putting the blame of every WML failure on Lua...

How can you expect [store_unit] to store several units (!) into the single variable foo[1]? As zookeeper already explained, there are no multidimensional arrays in WML. So this can't work and this doesn't have anything to do Lua (and this is actually caused by your code written in 2007, e.g. commit 19367).

It is counter-intuitive that [store_unit] cannot store a single unit to a single array position. I'd even consider it a bug. It needs to support more intelligent array operations, much like [set_variables].

While Lua can't be blamed for everything of course, it was a HUGE RECENT CHANGE and therefore I expect it introduced a fair number of bugs. If it did not, then I will be pleasantly surprised. However, it already has introduced bugs which were detected and presumably fixed, if I recall correctly. My nervousness as we approach 1.8 is nothing personal against you, I promise! I'm just scared of big changes and that is my nature I guess.

In this case, the bug is due to my failure to anticipate this particular usage.

However, I feel that storing to subarrays such as variable=foo1.foo2 should work in the current stable and dev versions. And this was something that did not work before my changes to [store_unit] in 2007. So that is mainly what I wanted to respond to, where zookeeper wrote it would not work, and I say yes it should. Going back over his words now, it seems that by "sub-array" he meant an explicit index of an array, so perhaps I misunderstood him.

[silene]

While Lua can't be blamed for everything of course, it was a HUGE RECENT CHANGE and therefore I expect it introduced a fair number of bugs. If it did not, then I will be pleasantly surprised. However, it already has introduced bugs which were detected and presumably fixed, if I recall correctly.

Let's put things into perspective a bit. Lua support is just 2'000 lines isolated in two C++ files. If you want "huge recent change", look at the AI instead, which went from 6'000 lines to 20'000 lines during the 1.7 cycle and which now spans 84 files. Now, what about bugs? On the top of my head, I can think of only two noticeable ones that were caused by the Lua support (dwarf leaders having 1hp and some menu options behaving badly). I don't doubt that other bugs will appear, but for now it is presumably one of the "huge changes" of Wesnoth that has experienced the lowest amount of bugs.

[Xudo]

I don't know exactly, but Lua is a interface between developer and Game. WML too.
Lua can have influence on WML, just because it can call to some WML functions/variables/libraries... with new commands. here can be conflict between them. AI just became larger. I don't say, what he haven't bugs, but storing units and choosing target for attack have difference. [acronym="It's my opinion"]IMO[/acronym]

Also want to know bug or not.
I think, what this code

Spoiler:

Code: Select all

[event]
	name=moveto
	first_time_only=no
	[if]
		{VAR_EQ unit.x 18}
		{VAR_EQ unit.y 17}
		[or]
			{VAR_EQ unit.x 19}
			{VAR_EQ unit.y 18}
		[/or]
		[or]
			{VAR_EQ unit.x 20}
			{VAR_EQ unit.y 18}
		[/or]
		[then]
		[/then]
		[else]
#action
		[/else]
	[/if]
[/event]

#define VAR_EQ VAR VALUE
[variable]
	name={VAR}
	equals={VALUE}
[/variable]
#enddef

same as

Spoiler:

Code: Select all

[event]
	name=moveto
	first_time_only=no
	[if]
		{VAR_NOT_EQ unit.x 18}
		{VAR_NOT_EQ unit.y 17}
		[or]
			{VAR_NOT_EQ unit.x 19}
			{VAR_NOT_EQ unit.y 18}
		[/or]
		[or]
			{VAR_NOT_EQ unit.x 20}
			{VAR_NOT_EQ unit.y 18}
		[/or]
		[then]
#action
		[/then]
		[else]
		[/else]
	[/if]
[/event]

#define VAR_NOT_EQ VAR VALUE
[variable]
	name={VAR}
	not_equals={VALUE}
[/variable]
#enddef

But WML think not. Who right?

[silene]

But WML think not. Who right?

WML is right. Fill the truth table of your logical condition if you are not convinced.

[Anonymissimus]

@xudojnik
Note that what you want can probably be achieved more easily by simply filtering with a comma separated list (which also creates the or-condition that you need):

Code: Select all

[event]
	name=moveto
	[filter]
		x=18,19,20
		y=17,18,18
	[/filter]
	#...
[/event]

@(somehow) topic:
However, I know that in 1.6, storing a unit to the variables of another unit works:
[store_unit]variable=unit_var
...
[store_unit]variable=unit_var.variables.another_unit_var
...

This should be the same like unit_var[0].variables.another_unit_var, although in my actual code the filter for the second store_unit is built in such a way that there can be only a single matching unit (filtering by id).

[silene]

@(somehow) topic:
However, I know that in 1.6, storing a unit to the variables of another unit works:
[store_unit]variable=unit_var.variables.another_unit_var

Yes, the code above is supposed to work, even in 1.8. What doesn't work is unit_var.variables.another_unit_var[17] (notice the index at the end).

[Xudo]

I check conditional variable before "filtering", but you give me nice idea... Now I make my code more simple to read.

[Xudo]

Now I face "20091213 20:34:37 error engine: failed to auto-store $unit at (-999,-999)" in stderr.txt (65MB of this strings)
I think, what this message appear, because some event haven't attributes x and y. But I have no idea where and what can be broken. Or there are some infinity cycle with broken action.
Someone face this error before? If you solve this problem, tell me where can I search my mistake, or tell me how do you fix this.