exploration 0.6.1

Tool for reading, formatting, and manipulating exploration graphs, which are graphs (or graph-sequences) that describe exploration of a game space, principally designed with Metroidvania games in mind.

Exploration

Overview

This program provides data types for representing the exploration of spaces that can defined (or abstracted) in terms of discrete decisions, such as videogame levels with multiple rooms and also other things like conversation trees or a city block grid.

It represents space using a DecisionGraph, which is a multi-di-graph indicating the transition(s) between decisions, which can include information about prerequisites for transitions as well as effects a transition might have on the world. There is also a convention for representing unexplored regions using specially-tagged nodes of the graph. An Exploration is a sequence of DecisionGraphs, along with a sequence of decisions indicating where the explorer was at each step, a sequence of transitions indicating which transition was taken at each step, and a sequence of states indicating extra state at each step. These representations were developed with Metroidvania games in mind.

Core capabilities include:

• Representing exploration processes as a series of decisions including partial information about not-yet-explored decisions.
• Creating maps and explorations from various text formats, including exploration journal formats.
• Reasoning about reachability modulo transition requirements in terms of powers that must be possessed and/or tokens that must be spent for a transition. TODO
• The ability to represent fairly sophisticated game logic in the DecisionGraph, and even construct playable maps. Game logic that can't be captured this way can still be represented through making custom changes to maps between exploration steps. TODO

Dependencies:

• Python version 3.8+
• networkx For underlying graph structures.
• pytest for testing, install with [test] option to get it automatically.

Installing

Just run pip install exploration. The egtool script should be installed along with the module.

You can then run python -m exploration.tests to run tests.

Getting Started:

The egtool script provides a command-line interface to core functionality. The exploration.main module provides equivalent entry points from Python. exploration.core provides the main types and explains how they fit together.

Plans

• Better support for open-world games, where decisions are not as closely linked to virtual space structure.

Changelog

• v0.6.1 fixes egtool.py script to include new optional command-line args instead of mandatory args.
• v0.6.0 adds Graphviz Dot format support (both import and export) and fixes up some bugs with the JSON import/export. Also changes tags so that they have values instead of just being part of a set. Analysis tools have expanded a bit and a system of metafunctions for automatically applying smallest-unit analysis tools to larger units has been added so we don't have to re-write "apply this to each situation" a bunch of times. Analysis tools are set up to run every tool on a file and write results to a CSV file (see main.py for analysis tool configuration). The egtool.py script now accepts command-line options but falls back to interactive prompting. Finally finished describeProgress and overhauled analysis tests so that we don't have any xfails right now. Adds 'extinguish' and 'complicate' journal entry types to better support mistaken impressions.
• v0.5.1 reorganizes some testing code out of the import-module run path.
• v0.5 includes a few bugfixes over 0.4, and most notably a stable syntax for entering relative mode at the current location, as well as an 'F' entry type for "fulfills" to note power equivalence. It also introduces the first real analysis tool, which just counts the number of unexplored options at each step of the graph.
• v0.4 introduces the 'actionPart' target type so that you can do 'oa' to observe an action without taking it. It also introduces long-form entry types and targets: you can just write out the full name of an entry type, possibly followed by an @ and either a full or abbreviated target type. Hopefully this helps make things more accessible for beginners. There are also now a few built-in debug commands available for printing relvant stuff. More may be added as they become popular. It also introduces equivalences, stored in the DecisionGraph, which allow powers (but not tokens) to count as being obtained when one of a set of other requirements is met instead. The 'fulfills' journal entry can add these.
• v0.3 is a pretty big overhaul. Changes journal format a bit (zones are now easier to deal with). Adds command-line interface (__main__.py via API in main.py). Adds interactive script egtool. Adds JSON serialization and __eq__ methods for DecisionGraph and Exploration types. Fills out 'edit' effect type with a new Command syntax which is it's own tiny DSL for editing graphs/explorations. Also adds aliases in the journal format so that it's easier to store/recall complex but repeated patterns. Extends testing code quite a bit and fixes a fair number of bugs found via those tests, although coverage is still incomplete. This version pushes module compatibility up to 3.10 since the type checking code is to tightly intertwined with the actual runtime code in places to be easily separable and it uses 3.10 features. One final big change: the default journal format has changed so that 'g' is for tag, 't' (which used to be tag) is for 'retrace', and 'r' (which used to be retrace) is for 'return'). 'R' is no longer used by the default format. Note that until 1.0, there may continue to be some instability in the default journal format.
• v0.2 Journal functionality (journal.py) is working at a basic level, with a few things still to-do (e.g., edit effects). Design has changed since previous versions. The zone system now works, although needs more testing. Design will be iterated on so the API and particularly the journal format is not 100% stable yet. Changed version support from 3.7+ to 3.8+ because of needing typing.Literal.
• v0.1.2 Core functionality (core.py) is working & tested, with the exception of the zones system. Journals are not working, and tests for those have been disabled for now. Could be used for representation purposes, but is not yet complete. This version is effectively the first alpha release since I'm demoing at the PCG workshop. Note most of the 'core capabilities' are still TODO.
• v0.1.1 Still pre-alpha as it's in the process of being re-architected a bit, but some core functionality is present if rough (e.g., core.DecisionGraph and core.Exploration).
• v0.1 Initial pre-alpha upload.