June 9, 2015
Spent last couple days playing Dwarf Fortress on the go (too bad there’s no save button yet). This resulted in fixing some crashes, making UI/UX improvements related to repeating operations like placing buildings, and also dramatically improving network performance and responsiveness. Here’s a video demonstrating gameplay over a 3G cellular connection to a server in another country (175ms ping). Actually, it’s a bit smoother in real life then on a recording.

I’m now finishing this stuff, then will be another round of implementing missing screens and functionality, and then I’ll prepare an alpha version to give out to number of people to try.

Experimenting with UI colours again

June 5, 2015
Dark background is required to be able to use standard DF colours for professions, jobs, materials and everything else, but initially I was going to use all the default colours for UI components. Probably only some screens will be dark, and some will be partly dark, as the item and unit info screens.

Also making row height smaller on certain screens either as an optional compact mode, or more likely not optional, because it’s much less comfortable to navigate long lists (units, trade goods) when only several items fit on screen, compared to one text line per item in DF on desktop.
June 5, 2015
Artifact list, item info screen, look mode, look list, container contents view. Look mode eventually will replace building query mode and unit selection modes because there’s no sidebar on mobile so no real differences between them anyway.
May 27, 2015
A post was planned here about how I rewrote a widely used job management and cost recovery software for MFDs from 450k lines in Java to 15k lines in Node.js, making it ten times faster and more reliable. But what else can I say?

Overcoming 16-colour palette limitations in Dwarf Fortress

May 19, 2015
In Text Will Be Text version 5.46 I’ve added a way to use colour values from colour tokens for tiles. If any background or foreground colour is >= 100, then colour values will be taken from df.global.world.raws.language.colors[value-100]. One example of how to use this is the included script realcolors.lua, which processes material definitions and replaces tile colours with values from material.state_color.Solid, which will give results similar to the screenshots below.

Before: After (note the different colours of wooden furniture): Technically this was very easy to do but some work probably will be required from tileset authors. Some colours specified in vanilla raws are not quite correct, but this can be fixed (e.g. by using an external colour database, one option is to use Stonesense data, as Japa suggested). The biggest problem here is that tilesets are designed to make use of background and foreground colours, while if we make use of colour tokens, there will be only one foreground colour. For the background colour, we’ll have to either leave it as is, or set to some fixed value (e.g. black or white), and that’s not what existing tilesets expect. EDIT: On the other hand, if you specify colours manually (in raws or overrides.txt for overrides), you can still have both foreground and background colours set to anything you want.
May 14, 2015
Will never understand companies spending time and money on copy protection, even developing seemingly sophisticated frameworks, that can be bypassed by patching several bytes. Hello, @fournova.
May 2, 2015

Here are some details on feedback to user actions and on behaviour in case of network issues.

The problem. There are basically three types of actions. Some you want to perform/queue quickly, e.g. marking many items for sale. You still want some feedback, but it shouldn’t prevent you from performing further actions. Some, on the other hand, should block the UI until they are executed and response received, e.g. actually selling items. Also, map-related actions do not require any feedback - when you change z-level you naturally wait for the screen to update.

I wanted the user experience to be smooth in case of possible frequent disconnections, either because of network problems, or because of switching to other apps, locking the phone, etc. So the reconnection must be quick and shouldn’t involve reloading more of the game state than required. At the same time, the game state must be always synchronised between the server and the app - if you issued a command to sell items, then disconnected and reconnected, you should see the result, and the command must not be sent/executed again.

The solution. The app and the server now maintain a queue of undelivered commands and responses that is preserved between connections to the same game. This means that no use action can ever fail because of a network problem - the app will try to reconnect until it succeeds and sends/receives all queued commands and responses, or until you cancel. You do not need to retry any actions manually. Implementation of game screens became easier too, there’s no need for any screen-specific actions to handle network problems and/or to synchronise the state after reconnection.

The video at the top shows how it all works from the user’s point of view. Timeouts are deliberately inserted in the server code, and also the same command was used to demonstrate both non-blocking and blocking behaviour. There’s a small set of commands that are transmitted separately and not handled by this mechanism, mostly related to map movement and updates.


Status update

April 7, 2015
Long time no updates, been busy with other things lately. Now finishing server management, game loading and all related screens. Also experimenting with and polishing networking code - spent some time playing on remote servers via real network connections (as opposite to all on local machine). With reasonable networks connection and server location (i.e. not on the other end of the world) everything seems to be playable fine.

Dwarf Fortress on iPad

March 28, 2015
Dwarf Fortress Remote on iPad is terrific. Many parts of the UI will need to be reworked though - more cool things are possible on a bigger screen and some existing elements are less comfortable. But iPhone is still the main focus - its UI will work on iPad even if suboptimal, but not the other way round.