SoC2011 Student Page Aethaeryn

From The Battle for Wesnoth Wiki
Revision as of 19:46, 14 April 2011 by Aethaeryn (talk | contribs) (gui_message{}: whoops, misread the [message] specifications a bit)


This page is related to Summer of Code 2011
See the list of Summer of Code 2011 Ideas



This is a Summer of Code 2011 student page
Project: SoC_Ideas_Your_Own_Ideas2011



Description

Michael Babich - Improving Lua's User Interface Capabilities

I intend to improve the ability for add-ons to customize the user interface through Lua. I will focus on providing simple, but powerful syntax for a set of functions that will enable add-on creators to have improved, easier access to features that already exist within the game engine so that they are free to creatively provide more innovative content.

The purpose of this project is to act as a middleman between the game engine and the add-on creators using the Lua scripting language so that simple and powerful add-on creation can help revolutionize future MP, just as [set_menu_item] and [option] allowed for a great evolution in the quality of add-ons since 1.3. This interface framework needs to be as broad as possible in the time given and the scope allowed so that creative add-on authors themselves can stretch Wesnoth content far beyond its traditional style, but it also must provide a simplified Lua helper library so that everyone can use it without necessarily having to delve deeply into the complexity that exists in the currently-existing GUI.

IRC

Aethaeryn

Introduction

I have been making content for Wesnoth since 2007 and I have been gradually porting much of my more advanced content to Lua since late 2009. While doing so, I have noticed certain limitations to the current implementation of Lua as it relates to add-on creation, especially with certain elements only possible through WML or the rather poorly readable WML tables. As the title suggests, I wish to simply and extend the add-on creation capabilities in Lua. In particular, I will be working with various elements of the interface code because this is currently where Lua is weak and often just relying on WML tables. There are several problems with using WML tables in these areas:

  • Poor readability, clumsy syntax, and verbosity. The metatable in the helper library helps with readability, but does not always work (for instance on [or] and [and] boxes, where those words are reserved in Lua) and it still hides the underlying issue that true native Lua support is not there. It also is fairly difficult in certain applications where the XML-style WML syntax is clearly inferior and much more verbose than something done in pure Lua.
  • You're essentially using WML. This defeats the purpose of using Lua in the first place if the majority of your code in a particular area is in WML. This is made even more severe when one realizes that many of the areas that must currently be done via WML tables are complex things that are the kind of things best done in a real language in the first place. In particular, the [message], [option], and [set_menu_item] have no real Lua equivalent.
  • Calling Lua from WML tables is hacky. Currently, if one wants to call a Lua function from T.option {}, one has to set a string that is a call on a global Lua function.

An ideal scenario or campaign would be one that uses WML to define the basic structure and then call on Lua for any more advanced elements. As I spend most of my Wesnoth time making -- or playing -- content for MP that do things that WML was not originally intended to do, such as making Wesnoth an RPG instead of a TBS, I have noticed that WML does not scale very well as a general purpose language doing advanced scripting. Any traditional language, such as Lua, can handle certain advanced things in a much more readable and efficient manner, but there are certain gaps in Lua where WML tables must be integrated with the Lua, which defeats many of the primary reasons to use Lua in the first place.

Fortunately, the solution is rather obvious to the WML table problem. Lua needs to be extended to access features that almost all already exist in the game engine and these must be implemented through Lua functions and tables instead of WML tables so that the full benefits of using Lua for scripting can be realized. Although the problem and solution are fairly easy to describe, certain precautions must be taken to make ensure that save games are not broken and no new issues are introduced. In addition, the individual proposed changes must be enumerated for clarity as to the scope of the project.

The Implementation: Overview

The information can roughly be broken down into three separate phases, 0 through 2. First, I shall go through an overview so as to be straightforward with the overall planned changes, then I shall provide the details of each individual section, and finally I shall propose a detailed timeline as to when each part can be accomplished.

Phase 0: The Initial Changes

First, src/scripting/lua.cpp should be split into smaller, logically-organized files so that the file does not grow too large and unmanageable with the improvements planned. This has the additional effect of allowing me to get familiar with the core Lua code.

Next, various UI tags shall be implemented in native Lua without WML tables. The hardest portion of this is ensuring that the Lua changes can be implemented without breaking save files. Most of the existing WML code can be reused, so a key purpose of this is to get rid of the need for WML tables for this portion of the code so that it better interacts with files written purely in Lua and so it is in a consistent style with code in the later portion of this project. The trickiest part will be implementing the callback, which will allow menu options to trigger Lua functions.

Because the code already exists to handle the WML tags, most of the work is allowing these Lua functions to utilize already-existing code in the engine. In this phase, the Lua functions to be created more or less map one-to-one with WML tags, which is why this would be a "Phase 0". On the surface, the changes may seem subtle, but it will allow for a more Lua way of implementing these than WML tables, which will improve interaction with Lua scripting functions while also making the Lua side of Wesnoth scripting easier to learn. In addition, it will ensure that these very central UI actions are consistent with Phase 1 syntax.

The related WML tags that need to be ported to Lua functions:

  • [message]
    • [option]
    • [text_input]
  • [set_menu_item]

Phase 1: Complex Interface Changes

  • Implement improved Lua access to the in-scenario GUI stuff, again without WML table hacks.
    • This includes improving the syntax and implementation of the dialog Lua.
  • A way to take in text commands in a way similar to debug mode's :lua, but only using specific commands to do specific actions as defined by Lua code in the add-on.

Phase 2: Simplify Add-on Creation

Phase 0 and Phase 1 might seem somewhat unrelated, but Phase 2 is intended to unify them under a common, easily-accessible package of sorts that provides a highly-abstracted, clean syntax for add-on authors to manipulate various elements of the UI. The original implementation from Phases 0 and 1 will still exist for those who want more control and customization, while Phase 2 will make things available for those who do not wish to necessarily deal with the complexity. There are two areas that will allow the improvements to be immediately accessible in an easy manner to content authors:

  • I will implement a Lua "wesnoth.ui" that adds an additional abstraction layer on top of the already existing syntax, simplifying the appearance of the code at the expense of some finer control.
  • If I have the time, I will create new, defined-in-Lua WML tags that utilize wesnoth.ui so that one does not have to use Lua to benefit from the improvements to Lua.

The Implementation: Details

Phase 0

The goal of this phase is to implement similar behavior in Lua to important UI-related WML tags so that these elements can be used in Lua without WML tables, thus eliminating clumsy syntax and ensuring complete implementation. Currently, certain parts, including [option], require clumsy workarounds when implemented in WML tables due to the nature of the tag. This also sets the stage for the later phases by establishing a starting point for more complex feature additions.

Sample Syntax

local msg = gui_message{
   message  = "Hello world!",
   speaker = "narrator",
   image   = "wesnoth.png",
   options  = {
         { text='hello', callback = foobar() },
         { text='goodbye', callback = say_goodbye() }
   }
}

local msg2 = gui_message{
   message = "What's your name?",
   -- no "settings =" makes it set reasonable defaults
   -- text_input takes either a table containing the information, or a table of tables that are all valid text_input
   text_input = {variable = "foo", label = "Name: ", max_length = 42}
}

if side = host then
   msg.show

function foobar()
   msg2.show

gui_message{}

The basic portion of gui_message{} is an implementation of the [message] WML tag without any of the subtags. Through the extensions below it allows absolutely everything in WML [message] to be converted except for the filters, which are better done through plain Lua. Note that the use of {} instead of () for the function means that it technically is fed in a Lua keyed table, meaning that the order of the arguments does not matter as long as they match the key. Default values are used for arguments that are not defined, just like the WML behavior.

gui_message {
   message     = string,
   speaker     = string, -- "narrator", "unit" or "second_unit"
   side_for    = integer, -- or a table with a set of integers
   image       = string,
   caption     = string,
   scroll      = boolean,
   duration    = integer,
   sound       = string,
   options     = table, -- see below
   text_input  = table, -- see below
}

The best part about this is that the core function (excluding the options and text_input tables) can be easily converted into a form recognizable by WML and handled using the same function that handles the existing [message] code. Complexity and a difference from the WML definition only really enter when implementing [option] and [text_input].

Calling the function does not actually display the message. Instead, one does "variable = gui_message{ ... }" and then variable.show when one wants the message to trigger.

The alternative form of gui_message{} is gui_message(table) where table is a table with keys that match the definition of gui_message{}.

options

Options has two possible syntax. If there is only one option, it can directly contain the option table:

options = {
   text     = string,
   filter   = condition, -- must evaluate to true for the option to show
   callback = function, -- the actual effect of the command
}

If there is more than one option, it must have a table containing tables:

options = {
   {
      text     = string,
      filter   = condition, -- must evaluate to true for the option to show
      callback = function, -- the actual effect of the command
   }, {
      text     = string,
      filter   = condition, -- must evaluate to true for the option to show
      callback = function, -- the actual effect of the command
   },
}

callback

text_input

set_menu_item()

Phase 1

Dialog Improvements

GUI 2 Access

Optional: Commands

Graphical User Interfaces are great for most tasks, but there are situations where text-based input is a better way for commands to be run. If possible, it would greatly improve the capabilities of Lua if it were able to interpret chat messages. Lua already has the capability to output to chat, but input from chat would allow for Lua to become even more powerful in this area. Commands could be thought of in two different types, each with different use cases:

Public chat. With Lua given the capability to parse public MP chat, this allows for the computer side, the AI, to be able to interact with the players given certain keywords, which would give certain scenarios a more immersive feel. Instead of being merely cosmetic, this chat could have an actual effect on gameplay. For instance, certain taunts could anger the AI, causing it to change tactics to attack a different player or recruit different units.

Private commands with a /lua command. One is currently able to run Lua through :lua when debug mode is on, but this is infeasible for multiplayer. Being able to interpret "/lua command", especially if it were context aware for which hex is currently selected, would allow for several possibilities. For instance, allied AI could uses Lua to respond to your input so that it can formulate a strategy with you. Also, advanced content that could use certain /lua commands to trigger actions, allowing for anything from an MP-compatible "debug mode" to more restrictive triggers that are fitting to the particular scenario. Perhaps the host would use "/lua difficulty hard" to change the difficulty mid-scenario, or a player could use "/lua inventory" on their turn to pop up an RPG-style inventory screen without having to use any right click menus at all.

Essentially, the Lua would be able to use certain input as a trigger for executing Lua functions, enabling a far broader range of possibilities for content creators. This would not actually implement any changes to the way the game is played. Rather, it would provide the ability for others to use this new feature in their own scenarios, eras, AI, etc.

Phase 2

Sample syntax

The final goal is to use wesnoth.ui.show() to hide almost all of the interface details by providing a syntactical shortcut to the functions in Phase 0 and Phase 1. More complexity will be possible, as in sample2, but the extremely basic sample1 will be legal syntax as well.

sample1  = {message = "Hello world"}
sample2 = {message = {"Hello world", options = {speaker = "Konrad", side_for = 1}}}
wesnoth.ui.show(sample1)
wesnoth.ui.show(sample2)

This will allow for possibilities like the following, where the GUI message is literally as simple as providing one string:

function message_all(messageText)
   wesnoth.ui.show{message = messageText}

wesnoth.ui

WML additions

The Implementation: Timeline

now April 08 - April 22  

Pre-Summer I

Discuss details of the implementation with various developers in order to finalize most of the project's design.

Work on basic patches, especially in the early areas of Phase 0, so that commit access can be granted before the designated GSOC period.

April 23 - May 08

Continue to familiarize myself with the project and its code by doing more of Phase 0.

May 09 - May 18 Spring semester ends. Availability may be limited during this period.
May 19 - May 30

Summer I

Phase 0 ends; Phase 1 beings. My schedule is be entirely free in this time span.

May 31 - June 30

Summer II

The entire month of June will be spent on Phase 1. I will balance my time with a summer semester, but should have Fridays through Sundays entirely free.

July 4 - July 8 Expected busy week as my one summer course wraps up.
July 9 - July 17

Middle of GSOC

Phase 1 ends. The largest portion of the project, Phase 1, should be concluding around this point.

Since this is around the middle evaluation point, part of the focus should be on reviewing the design and plans for the second half of GSOC. If I am ahead of schedule or behind schedule, adjustments can be made.

July 18 - July 22 Summer semester ends. Expected busy week as my other summer course wraps up.
July 23 - August 15 

Summer III

Phase 2 begins. Again, my schedule should be entirely free to focus on the project.

August 15

Pencils Down Date

According to Google: "Take a week to scrub code, write tests, improve documentation, etc."

lim t
t→∞
Continue to work on Wesnoth's Lua scripting capabilities because it certainly will take more than just one summer to perfect it.

Goals

Patches

Currently, the Phase 0 patches are in progress.

Questionnaire

SoC Application

Battle for Wesnoth - Google Summer of Code Application