OpenKropki Activity
open-source implementation of the "kropki" (aka "tochki") game
Brought to you by:
mv_fox
Activity for OpenKropki
-
Mateusz Viste committed [r292]
minor optimizations
-
kropcore provides a rect of changed area on field after each move so UIs can redraw the field selectively
-
floodfill() replaced back with a stack-based approach since the array-based one was about 1000x slower
-
floodfill() fix
-
replaced floodfill() by a version that does not rely on stack
-
removed dead lines
-
internal bitfield modified so walls are marked with only one bit
-
replaced binary magic values by nice defines
-
moved okp import-export routines to okp.c
-
player names moved out of kropcore to keep it as small as possible
-
playablefieldx and playablefieldy are 8 bit now, for consistency with other members of the kropki_game struct
-
field dimensions are passed as unsigned char to save memory on tiny platforms
-
core files renamed to fit 8+3 filenames
-
ai_proto fix
-
adjusted tigran settings so it does not burn cpu that much
-
copyright date bump
-
duplicate wall searching finds also reversed wall (wall A-B is the same thing as wall B-A)
-
nicer ring to mark the dot last placed on playfield + colors are passed as uint32_t types
-
user language auto-detection
-
renamed LANG_LAST to LANG_ERR for future use as an error indicator
-
automated last-version detection
-
list tar.* and zip files in the archive
-
wording improvement
-
download is local + styling improvements + minor rewording
-
split addWall() with a separate findWall() routine for better readability and possible future reusage
-
OpenKropki released /v0.4/openkropki-0.4-win32.zip
-
-
freezed v0.4 to tags
-
fixed build documentation to reflect the filename change of the windows makefile
-
improved wording
-
hints are rendered with a bigger margin from the mouse cursor for better readability
-
styling improvements
-
improved wording
-
renamed Makefile.win to Makefile.mingw
-
preinit coordinates in case ui_menu() gets a list where first entry has RELPOS set
-
remove default case so I still get warnings about enum coverage in the future should new cases be added
-
added a default enum case to shut gcc warnings
-
initialize variables to shut gcc warnings
-
object files are declared only once for better consistency
-
replaced obsolete gettimeofday() by clock_gettime()
-
added missing headers so mingw does not complain
-
explicit array type
-
do not build genan - ai_proto is not used anyway
-
removed dead symlink
-
tigran does not discards fields with only 1 liberty
-
do not build genan - ai_proto is not used anyway
-
cosmetic code changes
-
improved translations
-
doc improvements
-
hints are computed using a common definition + added hints for language selection
-
tigran: early prunning of branches that yield poor results
-
debug flags are controlled from within the makefile
-
do not analyze fields that are almost captured by opponent, unless they lead to an immediate win
-
if no last move found, rely on g->playablefieldx,g->playablefieldy
-
fixed finding playable fields on almost-full games
-
inline (unsafe) KROPKI_GETPOS for faster queries when boundaries checks are not needed
-
configurable lastmove to add or not AI moves into list
-
AI-controlled fields highligthing for debug purposes
-
compute opp index once and reuse
-
added an AI reset to tigran
-
improved description, decreased fieldslookup and increased lastmoves count, also ignoring analysis of 0,0 corners
-
simpler election of interesting moves: choose fields moving spirally around last moves
-
fixed saving own moves in history array
-
tigran analyze lists of fields instead of looking at multiple speparate boxes
-
agregate analysis of same field for multiple boxes
-
cosmetic reorganization of the playconf menu
-
default opponent is Tigran now
-
note about new AIs
-
kropki_setpos() acceots a NO_WALLS flag to bypass walls creation, thus speeding up AI computations
-
ai_tigran chooses a random solution among all equally good solutions
-
seed the random generator so AIs can rely on rand()
-
tigran: do not analyze a 0,0 move - it is never usefull
-
tigran looks at last n moves + implemented multithreaded analysis
-
lots of Tigran fixes
-
added experimental AI Tigran
-
experimental minmax-like computations within karl
-
kropki_copygame() allows ignoring wall creation for faster processing of AI simulations
-
mcts field selection looks farther only when close fields are not available
-
proto AI replaced by Karl
-
mcts election looks at wins/loss counters + a single pass of immediate wins is computed
-
bugfixes
-
fixed counting (and locating) playable fields: a newly-captured field could be reported as "playable"
-
ai_karl draft syntax fixes
-
const-ified the game argument in two functions
-
ai_karl draft
-
debug log about loaded resume state contains some extra info
-
reset nextplayer in okp-loaded games that are ended
-
always scan playfield from top left to bottom right for consistency
-
take care to fill playablefields, playablefieldx and playablefieldy with proper values when initing a new game or loading a game from okp
-
kropki_setposforce() removed, functionally replaced by kropki_pass()
-
removed kropki_getpoints() for simpler API: app is expected to look directly into the game struct
-
optimization: calculating points no longer requires a playfield scan, points are precomputed during new-walls analysis
-
optimization: look for new walls only if new area got captured
-
optimized findfirstfreefield() so it looks for a precomputed position instead of scanning the field
-
added new AI prototype
-
constify argument of kropki_findfirstfreefield()
-
computing playable fields left in kropki game stats, may be useful for some progress stats indicator
-
translations of the Dumbo AI mouseover hint
-
added a hint system for menus, used right away for providing a short description of the Dumbo AI
-
okp resume state contains the types of players now
1 >
