supp-ran.lua makes code unable to work in ConTeXt
I've adapted the following code to make mazes in LuaLaTeX and ConTeXt: https://www.rosettacode.org/wiki/Maze_generation#Lua I defined a modified version to make content visible to TeX and I saved in a file named maze.lua: function make_maze_tex(w, h, m) w = w or 16 h = h or 8 local map = initialize_grid(w*2+1, h*2+1) function walk(x, y) map[y][x] = false local d = { 1, 2, 3, 4 } shuffle(d) for i, dirnum in ipairs(d) do local xx = x + dirs[dirnum].x local yy = y + dirs[dirnum].y if map[yy] and map[yy][xx] then map[avg(y, yy)][avg(x, xx)] = false walk(xx, yy) end end end walk(math.random(1, w)*2, math.random(1, h)*2) tex.print([[\bgroup]]) tex.print([[\baselineskip ]]..m) local s = {} for i = 1, h*2+1 do --table.insert(s, [[\hbox{]]) tex.print([[\hbox{]]) for j = 1, w*2+1 do -- table.insert(s, [[\hbox{]]) if (j == 1 and i == 2) or (j == 2*w + 1 and i == 2*h) then tex.print([[\hskip ]]..m) elseif map[i][j] then --table.insert(s, [[\vrule width ]]..m..[[ height ]]..m) tex.print([[\vrule width ]]..m..[[ height ]]..m) else --table.insert(s, [[\hskip ]]..m) tex.print([[\hskip ]]..m) end -- table.insert(s, [[}]].."\n") end --table.insert(s, [[}]]) tex.print([[}]]) end --tex.print(table.concat(s)) tex.print([[\egroup]]) end And I executed the folowing in LuaLaTeX and ConTeXt: %\documentclass{article} %\usepackage{luacode} %\begin{document} \starttext %\begin{luacode*} \startluacode mz = dofile("maze.lua") local count = 0 for i=10,109 do count = count + 1 tex.print([[\subject{Laberinto ]]..count..[[}]]) mz.make_maze_tex(i, i, [[\dimexpr\textwidth/]]..(2*i+1)..[[\relax]]) tex.print([[\pagebreak]]) end %\end{luacode*} \stopluacode %\end{document} \stoptext Whereas it works very fast in LuaLaTeX it crashes on ConTeXt and outputs the following: token call, execute: ...ext/tex/texmf-context/tex/context/base/mkiv/supp-ran.lua:30: C stack overflow Why does it happen? I guess it has something to do with math.randomseed, but I don't know how to avoid this error message.
On 5/1/2020 6:12 AM, Jairo A. del Rio wrote:
I've adapted the following code to make mazes in LuaLaTeX and ConTeXt: https://www.rosettacode.org/wiki/Maze_generation#Lua
I defined a modified version to make content visible to TeX and I saved in a file named maze.lua:
function make_maze_tex(w, h, m) w = w or 16 h = h or 8
local map = initialize_grid(w*2+1, h*2+1)
function walk(x, y) map[y][x] = false
local d = { 1, 2, 3, 4 } shuffle(d) for i, dirnum in ipairs(d) do local xx = x + dirs[dirnum].x local yy = y + dirs[dirnum].y if map[yy] and map[yy][xx] then map[avg(y, yy)][avg(x, xx)] = false walk(xx, yy) end end end
walk(math.random(1, w)*2, math.random(1, h)*2)
tex.print([[\bgroup]]) tex.print([[\baselineskip ]]..m) local s = {} for i = 1, h*2+1 do --table.insert(s, [[\hbox{]]) tex.print([[\hbox{]]) for j = 1, w*2+1 do -- table.insert(s, [[\hbox{]]) if (j == 1 and i == 2) or (j == 2*w + 1 and i == 2*h) then tex.print([[\hskip ]]..m) elseif map[i][j] then --table.insert(s, [[\vrule width ]]..m..[[ height ]]..m) tex.print([[\vrule width ]]..m..[[ height ]]..m) else --table.insert(s, [[\hskip ]]..m) tex.print([[\hskip ]]..m) end -- table.insert(s, [[}]].."\n") end --table.insert(s, [[}]]) tex.print([[}]]) end --tex.print(table.concat(s)) tex.print([[\egroup]]) end
And I executed the folowing in LuaLaTeX and ConTeXt:
%\documentclass{article} %\usepackage{luacode} %\begin{document} \starttext %\begin{luacode*} \startluacode mz = dofile("maze.lua") local count = 0 for i=10,109 do count = count + 1 tex.print([[\subject{Laberinto ]]..count..[[}]]) mz.make_maze_tex(i, i, [[\dimexpr\textwidth/]]..(2*i+1)..[[\relax]]) tex.print([[\pagebreak]]) end %\end{luacode*} \stopluacode %\end{document} \stoptext
Whereas it works very fast in LuaLaTeX it crashes on ConTeXt and outputs the following:
btw, fast is relative ... no problem making the code twice as fast it is actually a nice example for the cld manual ... attached some output from a first 'improvement' (that version runs some 50-60% faster so you have a challenge ... actually, it could be a nice challenge for others on this list: how to make this one contextish and so, maybe we should have a challenge every few weeks)
token call, execute: ...ext/tex/texmf-context/tex/context/base/mkiv/supp-ran.lua:30: C stack overflow
Why does it happen? I guess it has something to do with math.randomseed, but I don't know how to avoid this error message. no, it's just lua running out of stack space so it depends on the binary (and als the initial random value)
luatex 1.13 uses lua 5.3 which has a large stack but depending on the compiler/linker settings and platform (it's 1000000 but depending on the situatation you get a message or crash when that doesn't work out) luajittex has a limited stack of 8000 so it will also crash luametatex uses lua 5.4 which uses a bit different stack model and can't go that high (it has a default of 2000 but i will bnump that to 6000 which still seems to work ok, that way i get upto a 120 maze) (new lmtx upload later this weekend) Hans ----------------------------------------------------------------- Hans Hagen | PRAGMA ADE Ridderstraat 27 | 8061 GH Hasselt | The Netherlands tel: 038 477 53 69 | www.pragma-ade.nl | www.pragma-pod.nl -----------------------------------------------------------------
On Fri, 1 May 2020, Hans Hagen wrote:
On 5/1/2020 6:12 AM, Jairo A. del Rio wrote:
I've adapted the following code to make mazes in LuaLaTeX and ConTeXt: https://www.rosettacode.org/wiki/Maze_generation#Lua
I defined a modified version to make content visible to TeX and I saved in a file named maze.lua:
function make_maze_tex(w, h, m) w = w or 16 h = h or 8
local map = initialize_grid(w*2+1, h*2+1)
function walk(x, y) map[y][x] = false
local d = { 1, 2, 3, 4 } shuffle(d) for i, dirnum in ipairs(d) do local xx = x + dirs[dirnum].x local yy = y + dirs[dirnum].y if map[yy] and map[yy][xx] then map[avg(y, yy)][avg(x, xx)] = false walk(xx, yy) end end end
walk(math.random(1, w)*2, math.random(1, h)*2)
...ext/tex/texmf-context/tex/context/base/mkiv/supp-ran.lua:30: C stack overflow Why does it happen? I guess it has something to do with math.randomseed, but I don't know how to avoid this error message. no, it's just lua running out of stack space so it depends on the binary (and als the initial random value)
But why is it running out of stack? Isn't the walk function tail recursive, in which case it should not cause a stack overflow (https://www.lua.org/pil/6.3.html). Aditya
On 5/1/2020 11:38 PM, Aditya Mahajan wrote:
On Fri, 1 May 2020, Hans Hagen wrote:
On 5/1/2020 6:12 AM, Jairo A. del Rio wrote:
I've adapted the following code to make mazes in LuaLaTeX and ConTeXt: https://www.rosettacode.org/wiki/Maze_generation#Lua
I defined a modified version to make content visible to TeX and I saved in a file named maze.lua:
function make_maze_tex(w, h, m) w = w or 16 h = h or 8
local map = initialize_grid(w*2+1, h*2+1)
function walk(x, y) map[y][x] = false
local d = { 1, 2, 3, 4 } shuffle(d) for i, dirnum in ipairs(d) do local xx = x + dirs[dirnum].x local yy = y + dirs[dirnum].y if map[yy] and map[yy][xx] then map[avg(y, yy)][avg(x, xx)] = false walk(xx, yy) end end end
walk(math.random(1, w)*2, math.random(1, h)*2)
...ext/tex/texmf-context/tex/context/base/mkiv/supp-ran.lua:30: C stack overflow Why does it happen? I guess it has something to do with math.randomseed, but I don't know how to avoid this error message. no, it's just lua running out of stack space so it depends on the binary (and als the initial random value)
But why is it running out of stack? Isn't the walk function tail recursive, in which case it should not cause a stack overflow (https://www.lua.org/pil/6.3.html). not really tail recursive: that is only true when it's the last action and return is used and it nests very deep
function foo(...) ... return foo(...) end versus function foo(...) ... if something then foo(...) end end Hans ----------------------------------------------------------------- Hans Hagen | PRAGMA ADE Ridderstraat 27 | 8061 GH Hasselt | The Netherlands tel: 038 477 53 69 | www.pragma-ade.nl | www.pragma-pod.nl -----------------------------------------------------------------
On Sat, 2 May 2020, Hans Hagen wrote:
On 5/1/2020 11:38 PM, Aditya Mahajan wrote:
On Fri, 1 May 2020, Hans Hagen wrote:
On 5/1/2020 6:12 AM, Jairo A. del Rio wrote:
I've adapted the following code to make mazes in LuaLaTeX and ConTeXt: https://www.rosettacode.org/wiki/Maze_generation#Lua
I defined a modified version to make content visible to TeX and I saved in a file named maze.lua:
function make_maze_tex(w, h, m) w = w or 16 h = h or 8
local map = initialize_grid(w*2+1, h*2+1)
function walk(x, y) map[y][x] = false
local d = { 1, 2, 3, 4 } shuffle(d) for i, dirnum in ipairs(d) do local xx = x + dirs[dirnum].x local yy = y + dirs[dirnum].y if map[yy] and map[yy][xx] then map[avg(y, yy)][avg(x, xx)] = false walk(xx, yy) end end end
walk(math.random(1, w)*2, math.random(1, h)*2)
...ext/tex/texmf-context/tex/context/base/mkiv/supp-ran.lua:30: C stack overflow Why does it happen? I guess it has something to do with math.randomseed, but I don't know how to avoid this error message. no, it's just lua running out of stack space so it depends on the binary (and als the initial random value)
But why is it running out of stack? Isn't the walk function tail recursive, in which case it should not cause a stack overflow (https://www.lua.org/pil/6.3.html). not really tail recursive: that is only true when it's the last action and return is used and it nests very deep
function foo(...) ... return foo(...) end
versus
function foo(...) ... if something then foo(...) end end
Thanks for the clarification. If you or anyone else is going to play around with generating mazes, definitely take a look at http://www.mazesforprogrammers.com/ and the live demos here: https://weblog.jamisbuck.org/2011/2/7/maze-generation-algorithm-recap Aditya
If you or anyone else is going to play around with generating mazes, definitely take a look at http://www.mazesforprogrammers.com/
and the live demos here: https://weblog.jamisbuck.org/2011/2/7/maze-generation-algorithm-recap
A nice overview of the relationship with spanning tress (by the same author) http://www.jamisbuck.org/presentations/rubyconf2011/index.html And another set of visualizations https://bost.ocks.org/mike/algorithms/#maze-generation Aditya
participants (3)
-
Aditya Mahajan -
Hans Hagen -
Jairo A. del Rio