Pokemon navigation: the shared core for AutoStory by dayfine · Pull Request #114 · dayfine/Arduino-Source

dayfine · 2026-11-01T05:32:02Z

Depends on #45 (on #44 and #43; #43 and #44 are upstream as PokemonAutomation#1828).

BDSP and FRLG AutoStory each had their own grid, planner, walker and map format, so none of it could be shared or tested once. This adds the game-agnostic core: Pokemon/Navigation has the grid types, the leg planner, the leg walker with its stop hooks, the route log lines both walkers write, and Pokemon::MapData, the one JSON map schema both games' maps are written in (tiles in shared glyphs, elevation, warps that say how they are entered, objects, named points and flag sets); Pokemon/AutoStory has checkpoints as events, segments and the tile pose, with ScreenMachine/StepTrace in CommonTools. Navigation includes nothing from AutoStory or any game. Tested with the LegPlanner, LegWalker, MapData, RouteLog, StoryEvents and StorySegments suites in fork CI on Mac, Windows Clang and Ubuntu.

🤖 Generated with Claude Code

https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

std::function is copyable, so copying one silently forks a stateful lambda's state, and it can't hold move-only captures. std::move_only_function fixes both, but Apple Clang's libc++ doesn't have it yet. MoveOnlyFunction is an alias of std::move_only_function where the library defines it (GCC, MSVC) and a small fallback elsewhere (macOS), with copying deleted. Tested with four unit tests on fork CI (Mac runs the fallback; Ubuntu and Windows clang run the std branch). Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

FRLG and BDSP each built a grid planner for AutoStory, and the two copies share the same ideas but can't serve each other, because each planner is tied to its own map type. This adds the part both games plan on: the grid types (tile positions and directions) and a leg planner that asks the game's step model what one press does and prefers legs that end at a wall, tested on a small made-up game. The leg walker, ScreenMachine and TileGrid/MapSet live with their users in other PRs. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

BDSP's new-game intro runs as a closed loop over the screens it knows: read the screen, take the one step that screen calls for, read again, and treat an unexpected screen or a state that keeps coming back as the sign of a dropped or wrong input. This adds that loop as its own module: ScreenMachine runs it and StepTrace logs every recognized screen change with a numbered screenshot and every input, so a failure reports the screens that led up to it. Nothing here depends on navigation; there are no tests yet, CI checks that it builds on its own. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

AutoStory checkpoints are long C++ functions of walks, presses and waits, and a checkpoint only notices drift or a missed talk when a later step fails. This adds the game-neutral runner from the event-model design: a checkpoint is a list of events with skip_if, require, an action, expects, a learned pose and named recoveries (wait, then re-anchor, then stop), plus per-checkpoint interrupts. run_checkpoint() skips events whose effect already shows, so a reload resumes by re-checking, and it logs one "[Event] ..." line per event for the lab watcher. Observation and actuation are MoveOnlyFunctions, so the Pokemon::StoryEvents unit tests run it against a scripted world; no game uses it yet. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

dayfine · 2026-11-01T06:19:07Z

LGTM

Scarlet/Violet will run story checkpoints on map points and fly points, so the engine's Pose can no longer be tiles and a facing. The engine now keeps only the map, the certainty and an opaque Position the game supplies and the engine only prints; the tile pose moves to Pokemon_TilePose.h for the 2D games, and InterruptPose::INEXACT becomes WITHIN_SET. No behavior change: the log text for a tile pose is the same as before. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

Upstream merged MoveOnlyFunction as PokemonAutomation#1810, so this branch's copy is redundant: MoveOnlyFunction_Tests.cpp takes upstream's version, the header and test list were already identical. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

A story segment is now a Segment: rows of CheckpointSpec, each with an id, a title, whether the segment saves after it, and a factory that builds the checkpoint fresh for every attempt. run_segment() runs a range of rows in order through SegmentHooks (save, reload, log); a failed checkpoint is reloaded, rebuilt and run again up to SegmentPolicy's attempts, then the report names the checkpoint, the event and the reason. Actions no longer write the run's pose: a succeeded Outcome carries the pose and the runner writes it before the expects. describe() dumps a checkpoint or a segment as text for review, covered by a golden-string test. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

BDSP and FRLG each read their own map files with their own parser. This adds a game-neutral MapData and parse_map() in Pokemon/Navigation, with one schema: a name, the decompilation source (file, revision, date), tiles as rows of glyphs, named points, warps to a point of another map, objects (npc, item, sign, with a facing and a wander box), connections to the neighbouring maps, and patches of tiles measured on the console applied on top. The reader generalises FRLG's read_map(): errors name the source and the JSON path, and anything placed outside the tiles is an error. Grid<Cell> moves into Pokemon_Grid.h for it. No game reads these yet. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

GCC's -Wdangling-reference flagged the source object bound from as_object() with a temporary path string. Naming the path first silences it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

A step can be marked wild (the walk may be interrupted there, e.g. by a wild battle in tall grass). The planner tracks per step whether the leg in progress has a wild step, marks such legs wild, and adds LegCosts::per_open_wild_stop when one of them ends in the open: after an interruption a leg to a wall is simply walked again, while how far an open one got is a guess. The weight defaults to 0, so games that don't set it plan as before. Moved here from frlg/walk-interrupt-recount (b4d1976), which added it for FRLG's Route 1. Tests on the planner's toy grid: a leg through grass is wild, an open stop after grass is weighed, the weight needs a wild step, and a new leg starts out not wild. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

# Conflicts: # SerialPrograms/Source/ComputerPrograms/ComputerPrograms_UnitTestRunner.cpp # SerialPrograms/Source/Pokemon/Navigation/Pokemon_Navigation_Tests.h # SerialPrograms/cmake/SourceFiles.cmake

Brings in nav/shared (#114) through #117. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

…count Brings in nav/shared (#114) through #117 and #118, including the planner's wild-step weighting (7008f8c), which moved there from this branch's b4d1976. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

…-mvp #119 now has nav/shared (#114) through #117 and #118, so this PR's diff shows only its own files. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

LegPlanner's wild-step weighting moved from #119 to #114 (7008f8c) and merged down #117 → #118 → #119 → #120; #120's diff is PokemonFRLG only. Adds #119, #120, #121 to the tree and a file-ownership table. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

The shared engine replaces this branch's copies of the grid types, leg planner and story events. LegWalker goes: nothing in BDSP uses it, and FRLG brings its own on nav/shared. SandgemCenterEvents moves to the shared Pose (a Position) and returns learned poses in its Outcome. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E # Conflicts: # Common/Cpp/MoveOnlyFunction_Tests.cpp # SerialPrograms/Source/ComputerPrograms/ComputerPrograms_UnitTestRunner.cpp # SerialPrograms/Source/Pokemon/Navigation/Pokemon_Grid.cpp # SerialPrograms/Source/Pokemon/Navigation/Pokemon_Grid.h # SerialPrograms/Source/Pokemon/Navigation/Pokemon_LegPlanner.h # SerialPrograms/Source/Pokemon/Navigation/Pokemon_LegPlanner_Tests.cpp # SerialPrograms/Source/Pokemon/Navigation/Pokemon_Navigation_Tests.h # SerialPrograms/Source/Pokemon/Navigation/Pokemon_StoryEvents.cpp # SerialPrograms/Source/Pokemon/Navigation/Pokemon_StoryEvents.h # SerialPrograms/Source/Pokemon/Navigation/Pokemon_StoryEvents_Tests.cpp # SerialPrograms/cmake/SourceFiles.cmake

…p/maps-rederive Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

A leg's wild flag was worked out from the tiles on its start map only, so "north 22" from Pallet Town, through Route 1's grass beyond the map's edge, was walked without watching for battles (console B, 2026-11-01). It now comes from the planner's per-step wild flag (#114), set on tall grass and water. The wild-steps route test checks that leg. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

The leg walker and its tests, moved here from the FRLG PRs (#117 added it, #119 dropped the tile-at-a-time walk and added the stop hooks), so that the shared navigation code lives in one PR. Unchanged from #119's version: it walks one leg with a game's mover, clears interrupts on the way, and stops an open leg at an interrupt or a leg to a wall that is interrupted again at once. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

…mes from #114 # Conflicts: # SerialPrograms/Source/Pokemon/Navigation/Pokemon_LegWalker.cpp # SerialPrograms/Source/Pokemon/Navigation/Pokemon_LegWalker.h # SerialPrograms/Source/Pokemon/Navigation/Pokemon_LegWalker_Tests.cpp

…hared stack starts on #43-#45

…s PR Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

…base test) into nav/shared Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

Navigation is moving the player (grids, maps, legs, the walker, the route log); AutoStory is the story. StoryEvents, StorySegments and TilePose (the story's pose on a tile map) move unchanged to Pokemon/AutoStory, next to the select database and stats, with their tests registered from a new Pokemon_AutoStory_Tests.h. Navigation includes nothing from AutoStory. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

…y) into nav/story-steps # Conflicts: # SerialPrograms/Source/Pokemon/Navigation/Pokemon_Navigation_Tests.h # SerialPrograms/cmake/SourceFiles.cmake

StorySteps and its tests move unchanged to Pokemon/AutoStory, next to the events and segments they build on (#114), and register from Pokemon_AutoStory_Tests.h. The story run options include them from there. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

Grid is a plain value whose methods only read: a cell is changed through the free function cell_at(). The leg walker's mover is a callable (LegMover = std::function) instead of a one-method interface. Position keeps its virtual describe(), now with a default, since games extend it with their own data and read it back by type (TilePose). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

The shared LegMover is now a callable (#114): the D-pad hold is a plain function. Reading the player's facing at the start of a route moves into read_facing(), keeping walk_overworld_route() under the function size limit. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

…s logged The [Segment], [Checkpoint] and [Event] lines went only to the program's log, and an event whose game step threw left the runner with no "failed" line. SegmentHooks gains an optional status sink that gets every line again without its tag ("00.03/2 below_oaks_trigger: start"), for the video overlay and, once run records land (#132), for the run's events. An exception out of an event is logged as "[Event] ... failed: <what>" and "[Checkpoint] ... failed at <n> <event>: <what>", then rethrown. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

…/story-steps # Conflicts: # SerialPrograms/Source/Pokemon/AutoStory/Pokemon_StorySegments.cpp # SerialPrograms/Source/Pokemon/AutoStory/Pokemon_StorySegments.h

All optional in the JSON, so today's maps parse unchanged: - world: a map's place on a world grid shared with other maps. - regions: named, ordered tile lists (replacing numbered points). - tiles: ' ' off the game's map; '8' '2' '4' '6' one-sided walls. - elevation: rows of hex digits, the size of the tiles. - warps: how they are entered (walk_into, step_on, cave_entrance, walk_<direction> for mats), and a destination that may be left out for a map not loaded; warp_at() finds the warp on a tile. - objects: trainer sight and the ways a trainer looks; hide_flag. - triggers: coord-event boxes, with the story state they fire in. - flag_sets: flags set together, by name. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

…into nav/story-steps Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E # Conflicts: # SerialPrograms/Source/ComputerPrograms/ComputerPrograms_UnitTestRunner.cpp

dayfine and others added 4 commits October 31, 2026 21:59

dayfine added a commit that referenced this pull request Nov 1, 2026

AutoStory PR progress: collector 1 is #114

2f1fc45

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

dayfine added a commit that referenced this pull request Nov 1, 2026

AutoStory PR progress: #114 accepted, frlg/mvp started, BDSP merges

6db5c57

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

dayfine marked this pull request as ready for review November 1, 2026 06:19

dayfine added a commit that referenced this pull request Nov 1, 2026

Merge nav/shared (#114) at 19a0303 into base/bdsp-1717

9cc2003

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

dayfine mentioned this pull request Nov 1, 2026

FRLG overworld navigation MVP (collector 3) #117

Draft

dayfine and others added 2 commits November 1, 2026 00:12

dayfine added a commit that referenced this pull request Nov 1, 2026

Merge nav/shared (#114) at 6549f98 into base/bdsp-1717

f9c9ebe

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

dayfine mentioned this pull request Nov 1, 2026

BDSP AutoStory MVP v3: segments 00 and 01 as event tables, maps as JSON #121

Closed

dayfine mentioned this pull request Nov 1, 2026

BDSP navigation v4: story maps, planner and route records on MapData #127

Draft

dayfine added a commit that referenced this pull request Nov 1, 2026

Merge nav/shared (a38e457, #114) into nav/story-steps

93075ab

Merge autostory/r3-shared-stats (5718316, #45) into nav/shared: the s…

4991c08

…hared stack starts on #43-#45

dayfine added a commit that referenced this pull request Nov 1, 2026

AutoStory plan: linear shared stack #43-#45 -> #114 -> #122, BDSP map…

a1cf759

…s PR Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

dayfine changed the base branch from base/upstream-main to autostory/r3-shared-stats November 1, 2026 20:30

Merge autostory/r3-shared-stats (7cc66e3: #43 carries its select-data…

6c46f7c

…base test) into nav/shared Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

dayfine added a commit that referenced this pull request Nov 1, 2026

Merge nav/shared (a6a0920, #114: code health) into nav/story-steps

55accbc

dayfine added autostory AutoStory programs game:shared Shared or game-neutral code navigation Maps, planner, walker labels Nov 1, 2026

dayfine and others added 2 commits November 2, 2026 07:00

Merge origin/autostory/r3-shared-stats into nav/shared

3b96255

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E

Merge autostory/r3-shared-stats into nav/shared

7cdb95b

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AKYzhrMnPK61XHWEch4a6E # Conflicts: # SerialPrograms/Source/ComputerPrograms/ComputerPrograms_UnitTestRunner.cpp

Navigation Menu

Repository navigation

Pokemon navigation: the shared core for AutoStory - #114

Pokemon navigation: the shared core for AutoStory#114
dayfine wants to merge 22 commits into
autostory/r3-shared-statsfrom
nav/shared

dayfine commented Nov 1, 2026 •

edited

Loading

Uh oh!

dayfine commented Nov 1, 2026

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

dayfine commented Nov 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

dayfine commented Nov 1, 2026

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

dayfine commented Nov 1, 2026 •

edited

Loading