| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Another step towards scriptable @commands.
Reviewed-by: bjorn.
|
|
Another step for being able to make @commands scriptable
Reviewed-by: bjorn.
|
|
Step to be able to handle chatcommands by scripts.
Reviewed-by: bjorn.
|
|
Due to a wrong primary key, which covered only the state name, it was
impossible to use the same state name on different maps. This has now
been fixed.
Another problem was that the map variables were being included in the
global variables, because the related database query did not filter on
the map_id column properly. While fixing that, the map_id column now
allows explicitly marking a state variable as global (with the value 0)
or system variables (with the value -1).
System variables are currently not accessible from scripts, but that
could be changed later.
Reviewed-by: Yohann Ferreira
Reviewed-by: Erik Schilling
|
|
When a prepared SQL statement would have returned multiple rows, this
function would try to set the column header names multiple times which
throws the AlreadySetException.
Currently it doesn't seem that any prepared statements are meant to return
multiple rows.
Reviewed-by: Yohann Ferreira
|
|
Global functions 'warn', 'info', 'debug' as shortcuts to the respective
'log' call, which also support passing multiple parameters at the same
time, which will be separated by spaces.
Global tables 'map' and 'world' which provide convenient read/write
access to map and world state variables.
Reviewed-by: Yohann Ferreira
|
|
The AGMSG_REGISTER_RESPONSE message should be sent _after_ adding the
global world state variables to it rather than before.
Reviewed-by: Yohann Ferreira
|
|
TODO: Inform client about this change.
Reviewed-by: bjorn.
|
|
Forgot to do this in the previous commit.
|
|
|
|
Scripts mostly execute the Mana script API, and it seems like just
unnecessary verbosity to refer to the 'mana' table all the time. This table
no longer exists now.
Reviewed-by: Erik Schilling
|
|
When creating an NPC, you now provide its optional talk and update functions
directly rather than them being stored in a table on the Lua side and then
called in response to a global callback.
Also fixed an issue with a missing gender parameter to the delayed NPC
creation callback used by NPCs defined on the map (found by Erik while
reviewing this patch).
Reviewed-by: Erik Schilling
|
|
Reviewed-by: Erik Schilling
|
|
Reviewed-by: bjorn.
|
|
Rather than wrapping NPC functions up in coroutines in the Lua side, they
are now managed on the C++ side as "script threads", which are essentially
the same thing.
The main purpose is that the server can now know whether any of these long
running script interactions are still active, which will probably be useful
when adding the ability to reload scripts.
Reviewed-by: Erik Schilling
|
|
Reviewed-by: PjotrOrial.
|
|
This was meant to be amended to the previous commit...
|
|
The new function 'checkCurrentMap' will raise an error when no current map has
been set, eliminating the need to do custom error handling all over the place.
This also fixes several functions that would otherwise have simply crashed
when there was no current map.
Also cleaned up some "empty string parameter" checks.
Reviewed-by: Erik Schilling
|
|
|
|
Usage:
@effect <id>
@effect <id> <char>
@effect <id> <x> <y>
Reviewed-by: Bertram.
|
|
mana.map_get_pvp() now returns one of the constants in libmana-constants.lua
Reviewed-by: Bertram.
|
|
This allows the script to let the character perform a scripted attack but
the character still gets xp and killed monsters give drops.
You can now call:
mana.being_damage(target, dmg, dmg_delta, accurancy, type, element, source,
skill)
While on it I also added checks to the being_damage function.
Reviewed-by: bjorn, Bertram.
|
|
It's a bit embarrassing the way this has gotten out of hand. The error
checking was inconsistent, and in some cases wrong.
A host of new helper functions, starting with 'check' rather than 'get',
perform error handling on function arguments (they will not return when
encountering an error).
Reviewed-by: Erik Schilling
|
|
Without a timeout, trying to write to a locked database fails immediately.
Now it will only fail when it can't retrieve a write lock within a second.
Still, we need to go through all the SQLite code some time, since there
is a lot of error handling missing. We also need to make sure that errors
are handled correctly on the higher level.
Reviewed-by: Yohann Ferreira
|
|
The ScriptAction of the TriggerArea (which can be created by
mana.trigger_create) was still using a named global function for its
callback. Now it also uses a reference to a script function.
Since it was the last occurrence of a call to a global script function,
I've also removed the Script::prepare(std::string) overload.
Reviewed-by: Erik Schilling
Mantis-issue: 299
|
|
Item and monster classes could already be identified by either their id or
their name. Now the explicit values returned by 'get_item_class' and
'get_monster_class' can also be used as parameter.
In addition the above two getters learned to understand all three types of
parameter as well, rather than only supporting a name.
Reviewed-by: Erik Schilling
|
|
We have to be careful with introducing globals now that there is only a
single Lua state, so we shouldn't use globals unnecessarily. Any variable
should be declared 'local' unless there is a reason to make it global.
For additional safety we can also think about disallowing the use of globals
entirely. This also helps to catch typos in scripts.
Reviewed-by: Erik Schilling
|
|
Previously, global function names were defined in the respective XML
definitions of items, monsters and status effects. This was reasonable when
they all had the same state, but now they're sharing the single global
Lua state.
Now the Lua API provides access to the ItemClass, MonsterClass and
StatusEffect instances, on which callbacks for both standard and custom
events can be explicitly set.
Reviewed-by: Erik Schilling
|
|
Reviewed-by: bjorn.
|
|
This includes the quest reply, post reply, death notification and
remove notification.
Also, Script::Ref was changed from a typedef to a small class,
automating initialization and making the check for validness clearer.
Reviewed-by: Erik Schilling
|
|
Rather than relying on the availability of global functions with certain
predefined names, the Lua script now calls API functions to set which
function should be called on these global events.
This mechanism should make it easier to avoid name collisions in the global
namespace, which is important now that there is only a single script state.
For these global events this was not likely to become a problem, but this
solution can also be used for callbacks on specific item or monster types,
or even allow setting callbacks on certain instances.
Reviewed-by: Erik Schilling
Reviewed-by: Yohann Ferreira
|
|
Mainly introduced a helper function for getting to the Script instance for
the current Lua state.
Reviewed-by: Erik Schilling
Reviewed-by: Yohann Ferreira
|
|
Mostly removed bulky copyright headers and fixed indentation and line length.
Reviewed-by: Yohann Ferreira
Reviewed-by: Erik Schilling
|
|
No more Lua state for each status effect, monster, item effect or map. All
scripts are loaded into the same state. This should be more efficient overall
and make it easier to implement dynamic reloading of the scripts in the
future.
Now, this introduces the problem of name collisions between different Lua
scripts. For now this is solved by using more specific function names, like
'tick_plague' and 'tick_jump' rather than just 'tick'. The plan is however
to get rid of these globals, and register these callbacks from the script,
so that they can be local functions without the danger of colliding with
other scripts.
Reviewed-by: Erik Schilling
Reviewed-by: Yohann Ferreira
|
|
It's not nice to use ++ on an iterator that may be std::map::end(),
in my case this caused it to hang indefinitely.
Reviewed-by: Yohann Ferreira
|
|
Reviewed-by: Bertram.
|
|
Reviewed-by: bjorn.
|
|
I did some cleanup but i had a wrong path set inside qtcreator.
So i recompiled old versions and it all worked fine.
But when i recompiled from console i got a whole bunch of errors.
Sorry for pushing broken commits.
Reviewed-by: bjorn.
|
|
Moved a code comment about syncmessages to the bugtracker: #473.
Reviewed-by: bjorn.
|
|
Reviewed-by: bjorn.
|
|
It's easier to just talk about world data and to modify it as a whole.
If there is really a need to separate it, a project can still choose to do
that (and in whatever suitable way). There is no need to enforce this
separation or to do it in our example.
Reviewed-by: Erik Schilling
|
|
Reviewed-by: Erik Schilling
|
|
Just defined an inline function to make talking to the player easier.
Would be nice when NPC scripts could look more like this by default.
Also indented with 4 spaces, which we do in C++ so why not Lua files
(the file had mixed 2 and 4 space indentation before).
|
|
Based on a native Lua table with weak values, so that the user data objects
that are created can be garbage collected when no longer referenced.
Reviewed-by: Yohann Ferreira
|
|
Also read the gids as unsigned integers since that's how Tiled writes them
to prevent the number going negative when the highest flag is used.
Reviewed-by: Yohann Ferreira
|
|
This can now be done in attributes.xml through the
minimum and maximum attribute parameters.
I also changed the AttributeInfo struct as requested by bjorn.
Reviewed-by: Erik Schilling, Thorbjørn Lindeijer
|
|
Based on a templated helper class, MapObject references in Lua scripts are now
full user data objects. Using the '__index' member of their metatable, a
library is associated with it so that member functions can be called directly
on the object.
Reviewed-by: Yohann Ferreira
Reviewed-by: Erik Schilling
|
|
When creating a Lua table, it is possible to specify in advance how many
array elements and how many non-array elements this table will contain,
to avoid dynamic reallocations while filling the table. This wasn't used
optimally by the pushSTLContainer helper methods and some other cases.
Reviewed-by: Erik Schilling
Reviewed-by: Yohann Ferreira
|
|
Reviewed-by: Erik Schilling
|
|
luaL_checkstring might not return so the destructor of string will not be called.
Reviewed-by: Bjorn.
|
