How can I get current character's rol-rac-ali-gen combo for NetHackChallenge task?

[adyomin]

import gym  
import nle  
env = gym.make("NetHackChallenge-v0")  
print(env.character)  

Only prints "@".

Edit: If I understand correctly, text from taking action Command.ATTRIBUTES is not available via obs object atm.

[adyomin]

@condnsdmatters you mentioned this issue recently:

While I appreciate that determining the race and role of your player might be useful to you, if you are looking for a more direct route in to getting this info you should open a different issue and we can discuss there.

Is there a more direct way of getting this info than parsing Command.ATTRIBUTES output once you complete the #207?

[paulkent-um]

Hi, fellow NLE botter here. One thing you can do as a stop-gap measure is use the tty_chars element in the observation, which contains the raw characters that would be displayed on the screen if a human were playing. This isn't a practical solution to all use-cases of #207 (notably, reading off lists of multiple items on a single space is impractical in this manner), but it should be workable for reading off Command.ATTRIBUTES since there's no danger of dungeon tiles being misinterpreted as part of the pop-up.

I have found success using this sort of technique to digest the output of the list of visible monsters (accessible with action 93: whatis, then action 55: "M").

[heiner]

Hey -- would ; followed by . together with the alignment entry in the tty_chars work? E.g. GLANCE followed by WAIT? That gives a message like e.g. @ a human or elf (human tourist called heiner). Unfortunately, blstats doesn't contain alignment, which really is an oversight on our part (#238), but you can find "neutral" etc in the actual bottom line entry in tty_chars:

Heiner the Rambler                St:10 Dx:15 Co:15 In:13 Wi:10 Ch:13 Neutral
Dlvl:1 $:428 HP:10(10) Pw:2(2) AC:10 Xp:1/0 T:2

[adyomin]

would ; followed by . ... work?

It took me a moment to realize what this means!

Thank you both @heiner and @paulkent-um! Both approaches provide a way to get what I was looking for!