Spent all Saturday on Undo. Progress:

• Hotkeys Z and Y are bound to undo/redo by default. You can change it in Lix's options menu.
  
• All Undoables will now suggest what tiles shall be selected after apply/undo. In particular:
  When you undo a deletion, the undeleted tiles will be selected.
  When you duplicate tiles, the newly inserted tiles will be selected.
  
• Lots of refactoring towards immutability of the involved classes. But the asserts in the area where I considered triple dispatch have begun to fire when I duplicate a large selection of both gadgets and terrain tiles.
  

New to-do list:

• Fix above bug in the undoable copy-paste.
  
• Rotations and mirroring must become undoable.
  
• Z-ordering (move to foreground/background) must become undoable.
  
• Resizing of the map must become undoable, or clear the undo stack if we want to rush the feature's release.
  

Cleanup work that isn't necessary to release undo, but that I would like to tackle:

• Ditch the Hover class hierarchy in the editor. This is a parallel class hierarchy that wrap a selecte tile. Work more directly with the raw selection of tiles.
  
• Open a pull request against Phobos, the D standard library, to allow RedBlackTrees with immutable classes as payload.
  
• Open another pull request against Phobos to make SortedRange const-correct. I would have preferred SortedRange over RedBlackTree for the Lix code from the beginning, but RedBlackTree had fewer roadblock bugs.
  
• Attempt to fix a compiler bug: 21321: Class with unimplemented interface method compiles, links, then segfaults, if inherited through abstract base class. I ran into this yesterday. It would be the first time that I seriously look at the D compiler's source, it's a large and scary codebase.
  

^ Illustration of a pull request. You must be careful, watch for feedback, and don't mind the time spent. I had one pull already merged into Phobos a month ago.

In the end, running into 4 library/compiler issues within one big Lix feature shows what parts of D are rarely treaded: Heavy object-orientation, exhausting and combining the OO features of D, making stuff transitively immutable where possible, and using the more obscure areas of the standard library.

Crane would be delighted, if it weren't all so template-heavy.

-- Simon

I am biased because I have played Part 1 more than Part 2. Thus, I had to solve Part 2 from scratch, and it took longer than resolving Part 1.

Both levels have their red herrings, e.g.,
-- Simon

Talisman has 3 lines of text. The indentation of the 2nd line looks strange, I would left-align all three lines.

I wouldn't bold the talisman title. Reason: It's not more important than the remainder. (This is different from level titles in a level list.) If you want to make it sound more important than the other two lines, you can print the level title in grey instead of black. Hard-coding colors sucks though, people use dark themes etc. that rely on replacing default colors. Can you print the default color with 50 % opacity? >_>

Three lines eat lots of vertical space and requires lots of scrolling. Consider removing the line break between talisman title and level title?

When the talisman is a special save requirement, print the number of lemmings available, too; i.e., print: Save 30/30.

Wording: Talismans don't get unlocked, they're all available from the beginning. Rather, they are won/beaten/accomplished/achieved.

-- Simon

Fixed in 0.9.34.

-- Simon

As a version, you can write the datetime of the most recent modification into the level, and have the replay refer to that. Like the incremented integer, this would update even for cosmetic chances, but that might be even desirable: In Lemmings, you can't guarantee that a terrain change is purely cosmetic.

Version mismatch isn't important enough to warn the user every time. If the replay works, it's all good. If the replay fails, only then it's interesting to see if the replay expects the same level version as the level has.

The downside of the single incrementing integer (compared to datetime) is that one version can give birth to two subversions with different terrain (both by you, or by different people), but they would have identical versioning information.

-- Simon

Thanks!

Platformers don't turn on hitting terrain, yes, I believe this is a misdesign. Over 10 years ago, I copied this behavior from L2 because I assumed L2 to be the gold standard.

On a fresh design, I would have the platformer turn. It's inconsistent w.r.t. the builder, and (platformer and builder) are much more closely related than (miner and digger, where the digger won't turn on steel but the miner does).

---

Even with the turn-on-hit builder, a builder bridge is easy to make traversable in both ways. A two-way-walkable platformer bridge is much harder with a turn-on-hit platformer, unless the two shoresides are exactly the same height. The non-turning platformer gives us more leeway. But I doubt this counterargument wins against the inconsistency argument.

At least we can fix some backroutes by giving platformers instead of giving builders.

-- Simon

This game looks amazing, I'll definitely have to try this out sometime. [...] There's a set on eBay, I might get it to play at Christmas.

I warmly encourage you to buy it. The 2017 release of Zendo ("Zendo 2.0") is the current and recommended printing, and it often appears on UK/US Amazon or Ebay.

It comes with 9 pieces of each of the 9 kinds, enough to rarely run out. Also, it offers cards with lots of easier rules, to help novice masters not accidentally make a rule too obstuse and difficult. More experienced masters can still invent their own rules.

(The only other release is the green box from 2003, it's practically nowhere to be found nowadays. So no worries; as long as your box is the black-and-white square box, you have the 2017 printing.)

In theory, you could apply the principle to any game pieces, but the little colourful shapes make it particularly appealing.

you could probably do a thing maybe with lemmings levels or other things. We use to play it on this forum as well in the past.

Right, I too prefer Zendo with (either the 1.0 pyramids or the 2.0 shapes) over Zendo with letters, numbers, or playing cards. A few different colorful pieces combine in many different ways, they look amazing on the table and patterns pop out.

My older post on this, with reasoning.

Does caffeine do more damage than good, it probably does when we thing about it. Good on you Simon for giving that up for July time.

Are you still going cold turkey with it at the moment Simon?

Yes, I still haven't drunk any caffeine since.

Caffeine probably won't lead to permanent damage, maybe I'll come back to it in a couple of months. But its wakening power is strongest when one isn't habituated to it. Thus, I'd rather not have any caffeine until I really need it some evening.

-- Simon