Minecraft113

[gmcnew]

Fixes #1454!

For a future todo list, see this comment: #1454 (comment)

[agrif]

This is fantastic and quick work! I had another idea last night about how to avoid having to invent new block ids and make this all a bit easier to maintain, but I'd sooner get this merged and work on making it nice later.

The idea, FWIW: change the @material decorator to instead accept a block name and properties dict, then change it to automatically assign an internal block id to that combination. For old blocks before 1.13, have some extra info about how to translate from old block id + data into the new names, and from there into the new internal block id. The textures.py file would still construct an array mapping these internal ids to the block sprite to use.

At the end of it all, you could fairly quickly turn both the new paletted block data, and the old blockid + data system into an array of the new internal block ids, and then render using that. It would require only minimal changes to the C code, never requires maintainers to explicitly assign fake ids, and moves the name and properties of a block closer to its texture definition. It could also be changed only slightly to support mod blocks (if anybody still cares to do that).

Again: not something to worry about now. If nobody else takes a crack at it, I might. You've ton a huge bulk of the work here. I'll give it a test later this week and we can move forward.

[mide]

Do we have tests? I know there has been a "Test World" Overviewer has used in the past - with blocks of many types and orientations. That may help us verify everything looks good.

Existing Tests: overviewer/Minecraft-Overviewer/test

Edit: It looks like TravisCI may be doing that and you're trying to fix that now.

[eminence]

We have exmaple, which is our standard test world, but it's not 1.13 compatible. If someone wants to undertake converting it, that would be awesome.

(Actually, we'd probably like 2 version -- a 1.13 compatible version and a 1.12 compatible version)

[Plagiatus]

the win64 build works great on my windows PC. Is there a build for debian/ubuntu somewhere that I don't need to compile myself? or is there an ETA on an update to the package?
Love the work you guys are doing <3

[agrif]

@RedSparr0w Thanks for the PR, sorry it took so long to merge. It looks like it's mostly furnaces and dispensers left. Anybody know of any other blocks?

@Plagiatus no debian builds for this, sorry. If you really need them I can figure it out, but at the moment I'm scared a force-build will also install this branch in the debian repo.

[MrGraversen]

Really good job guys. Looking forward to this!

[Plagiatus]

@agrif i managed to build it myself in the meantime, don't worry.

a block i noticed missing is the quartz pillar.
and most stairs seem to always be in a default rotation and not in the one they actually have been placed in.

[matrixhacker]

Updates on this anyone?

[mtoensing]

Could somebody pls merge this? Or is this project finally dead?

[mtoensing]

@CounterPillow
Wouldn't it be a good idea to merge this 113 branch into master? I saw some threads on reddit and boards around the internet where ppl ask why Overviewer does not work with Minecraft anymore. If they are lucky somebody points them to the 113 branch but in 2019 this is the most recent version. I would say you archive a "112" branch for all the ppl how wont update their servers but the majority would get the most recent version of Overviewer and Minecraft. What do you say? =)

[enaut]

@mtoensing
So far this is considered a alpha or beta... so you can use it at your own risk but there are still known problems. - just my 2$ ;)

[CounterPillow]

I want to at least rewrite _packed_longarray_to_shorts in C or rewrite it in a less terrible way before merging this, since it is the main performance issue right now.

[mtoensing]

@CounterPillow

rewrite packed_longarray_to_shorts

Can we help? Is this a task that takes 4 months or 4 days? Because I believe that the "missing" support for 1.13 damages the reputation of overviewer more than some bugs.

[strmtrupr2]

I was able to pull this branch and it worked with mscs and for my 1.13 world! Thank you :)

[Plagiatus]

Miracles do happen occasionally

[CounterPillow]

Lies!

[agrif]

Ok, so I've merged this. Thank you everybody for your very hard work!

I believe there are a few extant blocks remaining. PRs for those will be merged soon. The remaining pain point is how slow the block unpacking code is. This could be improved by either rewriting it in C, or changing it to manipulate numpy arrays without iterating over them. I would prefer a numpy solution, but would not turn away either in a PR.

Again, thank you everybody for the work here. I never would have believed Overviewer would survive past 1.13, but here we are!

[mtoensing]

Woho!

[piegamesde]

The remaining pain point is how slow the block unpacking code is

May I point you to https://github.com/piegamesde/nbt/blob/develop/src/main/java/com/flowpowered/nbt/regionfile/Chunk.java#L231-L253? I don't know your performance requirements, but I haven't had any problems with this so far.

[agrif]

@piegamesde Thanks! but what we do now is essentially a version of that in Python, unrolled a little bit. The challenge is either rewriting it to use numpy arrays without any loops at all, or rewriting it in C and interfacing that C code to the python code.

And, of course, making sure the new code does the same thing as the old code.

[agrif]

I have rewritten the slow bit of code to be faster, but I want some verification it works for all of you before merging:

#1519

If anybody here could test it and report back (in the PR thread, please!) that would be very helpful.