Friday, June 17, 2011

Task description - contest starts now!

[Remark: This post was last updated on 10:00 June 17 Friday UTC. Search it for "Update" on your browser.]

Welcome to the ACM SIGPLAN ICFP Programming Contest 2011! The task this year is to write a program that plays the card game Lambda: The Gathering (LTG for short).

Rules

Each match of LTG is played by two programs (player 0 and player 1) in alternate turns (turn 1 of player 0, turn 1 of player 1, turn 2 of player 0, turn 2 of player 1, etc.). Each player owns 256 slots, numbered from 0 to 255, and a fixed set of cards. A slot consists of a field and an integer called vitality, ranging from -1 to 65535. A field holds a value, which is either an integer, ranging from 0 to 65535, or a function (in the sense of programming languages). A value is a valid slot number if it is an integer greater than or equal to 0, and less than or equal to 255. A slot is dead if its vitality is 0 or -1; otherwise the slot is alive. At the start of each match, every field is initialized to the identity function (a function that takes an argument x and returns x) and every vitality is initialized to 10000. The cards have fixed values as defined below.

In each turn, the player (called the proponent) performs either a left application or a right application. A left application applies (as a function) a card to the field of one of the proponent's slots. A right application applies (as a function) the field of one of the proponent's slots to a card. In either case, an error is raised if the card or the field to be applied is not a function, or if the slot is dead. The other player (called the opponent) is able to see which application the proponent has chosen on what card and slot. [Remark: In mathematics and programming languages (as well as in our rules), one applies a function to an argument, not vice versa; that is, "applying f to x" means "f(x)", not "x(f)".] [Remark: As a result of the rules, all function applications in LTG are "call by value": that is, the argument x of function application f(x) is always a value.]

The turn ends when an error is raised, when the number of function applications caused by the left or right application (including itself) exceeds 1000, or when the left or right application returns a value without exceeding this limit. In the last case, the field of the slot used for the left or right application is overwritten with the return value. In the other cases, it is overwritten with the identity function. Effects caused by function applications are not rewound and remain even if an error is raised or the application limit is exceeded. Cards are not consumed by applications and can be reused as many times as necessary.

A match ends after 100000 turns of each player, or when every slot of a player has become dead after a turn. In either case, a player wins if it has more slots alive than the other player. The players tie if the numbers of slots alive are equal.

Cards

The set of cards are fixed as follows. The effect of a card is undefined in any case not specified below. [Remark: The images are only for amusement and are not part of the official rules.]

Card "I" is the identity function. [Remark: It is called the I combinator and written λx.x in lambda-calculus.]

Card "zero" is an integer constant 0.

Card "succ" is a function that takes an argument n and returns n+1 if n<65535 (or returns 65535 if n=65535). It raises an error if n is not an integer.

Card "dbl" is a function that takes an argument n and returns n*2 (n times 2) if n<32768 (or returns 65535 if n>=32768). It raises an error if n is not an integer.

Card "get" is a function that takes an argument i and returns the value of the field of the ith slot of the proponent if the slot is alive. It raises an error if i is not a valid slot number or the slot is dead.

Card "put" is a function that takes an (unused) argument and returns the identity function.

Card "S" is a function that takes an argument f and returns another function, which (when applied) will take another argument g and return yet another function, which (when applied) will take yet another argument x, apply f to x obtaining a return value h (or raise an error if f is not a function), apply g to x obtaining another return value y (or raise an error if g is not a function), apply h to y obtaining yet another return value z (or raise an error if h is not a function), and return z. [Remark: The first function is called the S combinator and written λf.λg.λx.fx(gx) in lambda-calculus.]

Card "K" is a function that takes an argument x and returns another function, which (when applied) will take another (unused) argument y and return x. [Remark: The first function is called the K combinator and written λx.λy.x in lambda-calculus.]

Card "inc" is a function that takes an argument i, and

  • increases by 1 the vitality v of the ith slot of the proponent if v>0 and v<65535,
  • does nothing if v=65535 or v<=0, or
  • raises an error if i is not a valid slot number,

and returns the identity function.

Card "dec" is a function that takes an argument i, and

  • decreases by 1 the vitality v of the (255-i)th slot of the opponent if v>0,
  • does nothing if v<=0, or
  • raises an error if i is not a valid slot number,

and returns the identity function.

Card "attack" is a function that takes an argument i and returns another function, which (when applied) will take another argument j and return yet another function, which (when applied) will take yet another argument n, decrease by n the vitality v of the ith slot of the proponent (or raise an error if i is not a valid slot number, n is not an integer, or n is greater than v), and

  • decrease by n*9/10 (n times 9 divided by 10, with the remainder discarded) the vitality w of the (255-j)th slot of the opponent if it is alive (w is set to 0 if it would become less than 0 by this decrease),
  • do nothing if the slot is dead, or
  • raise an error if j is not a valid slot number,

and return the identity function.

Card "help" is a function that takes an argument i and returns another function, which (when applied) will take another argument j and return yet another function, which (when applied) will take yet another argument n, decrease by n the vitality v of the ith slot of the proponent (or raise an error if i is not a valid slot number, n is not an integer, or n is greater than v), and

  • increase by n*11/10 (n times 11 divided by 10, with the remainder discarded) the vitality w of the jth slot of the proponent if it is alive (w is set to 65535 if it would become greater than 65535 by this increase),
  • do nothing if the slot is dead, or
  • raise an error if j is not a valid slot number,

and return the identity function.

Card "copy" is a function that takes an argument i, and returns the value of the field of the ith slot of the opponent. It raises an error if i is not a valid slot number. Note that the slot is ith, not (255-i)th.

Card "revive" is a function that takes an argument i, sets to 1 the vitality v of the ith slot of the proponent if v<=0 (or does nothing if v>0), and returns the identity function. It raises an error if i is not a valid slot number.

Card "zombie" is a function that takes an argument i and returns another function, which (when applied) will take another argument x, and

  • overwrite with x the field of the (255-i)th slot of the opponent if the slot is dead, or
  • raise an error if i is not a valid slot number or the slot is alive,

and set the vitality of the slot to -1, and return the identity function.

Immediately before each turn of a player, the field of every slot of the player with vitality -1 is automatically applied (as a function) to the identity function (even though the slot is dead) in increasing order of slot number. [Update on 07:05 June 17 Friday UTC: The vitality of each slot is checked just before its automatic application; that is, (1) the vitality of slot 0 is checked, and (2) if it is -1, the field of slot 0 is applied, and then (3) the vitality of slot 1 is checked, and (4) if it is -1, the field of slot 1 is applied, etc.] For each of these automatic applications, an error is raised if the field is not a function or the number of function applications caused by the application exceeds 1000. An error caused by each automatic application does not affect any other automatic applications. After each automatic application, the field of the slot used for the application is overwritten with the identity function, and the vitality of the slot is reset to 0.

In addition, during the above automatic applications, parts of the effects of 4 cards change from their previous descriptions as follows (the other parts do not change from the original):

  • Card "inc" decreases the v (in the previous description of this card) by 1 if v>0, or does nothing if v<=0.
  • Card "dec" increases the v by 1 if v>0 and v<65535, or does nothing if v<=0 or v=65535.
  • The third function in card "attack" increases the w by n*9/10 if w>0 (w is set to 65535 if it would become greater than 65535 by this increase), or does nothing if w<=0.
  • The third function in card "help" decreases the w by n*11/10 if w>0 (w is set to 0 if it would become less than 0 by this decrease), or does nothing if w<=0.

[Remark: For an informational purpose only, executables for interactive LTG plays are provided at:

http://www.kb.ecei.tohoku.ac.jp/icfpc/ltg.linux32.gz
http://www.kb.ecei.tohoku.ac.jp/icfpc/ltg.linux64.gz
http://www.kb.ecei.tohoku.ac.jp/icfpc/ltg.win32.exe.zip
http://www.kb.ecei.tohoku.ac.jp/icfpc/ltg.macosx-10.4-ppc.gz (also for 10.5-ppc)
http://www.kb.ecei.tohoku.ac.jp/icfpc/ltg.macosx-10.5-intel.gz
http://www.kb.ecei.tohoku.ac.jp/icfpc/ltg.macosx-10.6-intel.gz

It is forbidden to use any of these executables as any part of your submission (such programs would be disqualified). Note also that these executables are not part of the official rules. The contest organizers do not provide any guarantee or support for them, although reports of bugs and problems are welcome if any. A simple example session (with alternate turns by two players) is shown below (details of the informational output may differ depending on versions). [Update on 10:00 June 17 Friday UTC: The numbers of turns shown in the previous version were incorrect. We will update the executables above as soon as possible. Although they are only informational and unsupported, we aplogize for such a stupid bug!]


$ ./ltg
Lambda: The Gathering version $Date:: 2011-06-17 18:39:42 +0900#$
Usage: ./native <options> only
         (one player; continues forever)
       ./native <options> alt
         (two players; continues forever)
       ./native <options> match prog1 prog2
         (similar to official match)

  -non-blocking <true|false>  Non-blocking I/O on pipes to prog1 and prog2
                              (default: false on Win32, true otherwise;
                               does not work on Win32)
  -silent <true|false>  Output only the last line (default: false)
  -print-fields <true|false>  Print fields (true, default) or not (false)
  -help  Display this list of options
  --help  Display this list of options
$ ./ltg alt
Lambda: The Gathering version $Date:: 2011-06-17 18:39:42 +0900#$
###### turn 1
*** player 0's turn, with slots:
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
2
slot no?
0
card name?
zero
player 0 applied slot 0 to card zero
*** player 1's turn, with slots:
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
2
slot no?
0
card name?
inc
player 1 applied slot 0 to card inc
###### turn 2
*** player 0's turn, with slots:
0={10000,zero}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
1
card name?
succ
slot no?
0
player 0 applied card succ to slot 0
*** player 1's turn, with slots:
0={10000,inc}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
2
slot no?
0
card name?
zero
player 1 applied slot 0 to card zero
###### turn 3
*** player 0's turn, with slots:
0={10000,1}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
1
card name?
succ
slot no?
0
player 0 applied card succ to slot 0
*** player 1's turn, with slots:
0={10001,I}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
2
slot no?
0
card name?
dec
player 1 applied slot 0 to card dec
###### turn 4
*** player 0's turn, with slots:
0={10000,2}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
1
card name?
dbl
slot no?
0
player 0 applied card dbl to slot 0
*** player 1's turn, with slots:
0={10001,dec}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
2
slot no?
0
card name?
zero
player 1 applied slot 0 to card zero
###### turn 5
*** player 0's turn, with slots:
0={10000,4}
255={9999,I}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
1
card name?
inc
slot no?
0
player 0 applied card inc to slot 0
*** player 1's turn, with slots:
0={10001,I}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
1
card name?
succ
slot no?
0
player 1 applied card succ to slot 0
Exception: Native.Error
slot 0 reset to I
###### turn 6
*** player 0's turn, with slots:
4={10001,I}
255={9999,I}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?

Another example session (with unilateral turns by a solitary player, which is not part of the official rules) follows, again for an informational purpose only. It shows how to compose (as functions) the field of a slot with a card, as well as how to apply (as a function) the field of a slot to that of another.


$ ./ltg only
Lambda: The Gathering version $Date:: 2011-06-17 18:39:42 +0900#$
###### turn 1
*** player 0's turn, with slots:
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
2
slot no?
0
card name?
help
player 0 applied slot 0 to card help
###### turn 2
*** player 0's turn, with slots:
0={10000,help}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
2
slot no?
0
card name?
zero
player 0 applied slot 0 to card zero
###### turn 3
*** player 0's turn, with slots:
0={10000,help(zero)}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
1
card name?
K
slot no?
0
player 0 applied card K to slot 0
###### turn 4
*** player 0's turn, with slots:
0={10000,K(help(zero))}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
1
card name?
S
slot no?
0
player 0 applied card S to slot 0
###### turn 5
*** player 0's turn, with slots:
0={10000,S(K(help(zero)))}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
2
slot no?
0
card name?
succ
player 0 applied slot 0 to card succ
###### turn 6
*** player 0's turn, with slots:
0={10000,S(K(help(zero)))(succ)}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
2
slot no?
0
card name?
zero
player 0 applied slot 0 to card zero
###### turn 7
*** player 0's turn, with slots:
0={10000,help(zero)(1)}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
2
slot no?
1
card name?
zero
player 0 applied slot 1 to card zero
###### turn 8
*** player 0's turn, with slots:
0={10000,help(zero)(1)}
1={10000,zero}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
1
card name?
succ
slot no?
1
player 0 applied card succ to slot 1
###### turn 9
*** player 0's turn, with slots:
0={10000,help(zero)(1)}
1={10000,1}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
1
card name?
dbl
slot no?
1
player 0 applied card dbl to slot 1
###### turn 10
*** player 0's turn, with slots:
0={10000,help(zero)(1)}
1={10000,2}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
1
card name?
dbl
slot no?
1
player 0 applied card dbl to slot 1
###### turn 11
*** player 0's turn, with slots:
0={10000,help(zero)(1)}
1={10000,4}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
1
card name?
dbl
slot no?
1
player 0 applied card dbl to slot 1
###### turn 12
*** player 0's turn, with slots:
0={10000,help(zero)(1)}
1={10000,8}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
1
card name?
dbl
slot no?
1
player 0 applied card dbl to slot 1
###### turn 13
*** player 0's turn, with slots:
0={10000,help(zero)(1)}
1={10000,16}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
1
card name?
K
slot no?
0
player 0 applied card K to slot 0
###### turn 14
*** player 0's turn, with slots:
0={10000,K(help(zero)(1))}
1={10000,16}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
1
card name?
S
slot no?
0
player 0 applied card S to slot 0
###### turn 15
*** player 0's turn, with slots:
0={10000,S(K(help(zero)(1)))}
1={10000,16}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
2
slot no?
0
card name?
get
player 0 applied slot 0 to card get
###### turn 16
*** player 0's turn, with slots:
0={10000,S(K(help(zero)(1)))(get)}
1={10000,16}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
1
card name?
K
slot no?
0
player 0 applied card K to slot 0
###### turn 17
*** player 0's turn, with slots:
0={10000,K(S(K(help(zero)(1)))(get))}
1={10000,16}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
1
card name?
S
slot no?
0
player 0 applied card S to slot 0
###### turn 18
*** player 0's turn, with slots:
0={10000,S(K(S(K(help(zero)(1)))(get)))}
1={10000,16}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
2
slot no?
0
card name?
succ
player 0 applied slot 0 to card succ
###### turn 19
*** player 0's turn, with slots:
0={10000,S(K(S(K(help(zero)(1)))(get)))(succ)}
1={10000,16}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
2
slot no?
0
card name?
zero
player 0 applied slot 0 to card zero
###### turn 20
*** player 0's turn, with slots:
0={9984,I}
1={10017,16}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?

The last example shows how to implement an infinite loop.


$ ./ltg only
Lambda: The Gathering version $Date:: 2011-06-17 18:39:42 +0900#$
###### turn 1
*** player 0's turn, with slots:
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
2
slot no?
0
card name?
S
player 0 applied slot 0 to card S
###### turn 2
*** player 0's turn, with slots:
0={10000,S}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
2
slot no?
0
card name?
get
player 0 applied slot 0 to card get
###### turn 3
*** player 0's turn, with slots:
0={10000,S(get)}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
2
slot no?
0
card name?
I
player 0 applied slot 0 to card I
###### turn 4
*** player 0's turn, with slots:
0={10000,S(get)(I)}
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?
2
slot no?
0
card name?
zero
player 0 applied slot 0 to card zero
Exception: Native.AppLimitExceeded
slot 0 reset to I
###### turn 5
*** player 0's turn, with slots:
(slots {10000,I} are omitted)
(1) apply card to slot, or (2) apply slot to card?

It is known that (closed) lambda-terms can be translated into function applications of S, K, and I (and I can also be implemented by S and K). A lambda-term M is either a variable x, a function application M(N) (often written MN without the parentheses), or a function λx.N that takes an argument x, evaluates another lambda-term N and returns its value. A translation T from (closed) lambda-terms to S, K, and I can be given as follows:

T(λx.M) = A(x,T(M))
T(MN) = T(M) T(N)
T(x) = x

where

A(x,x) = I
A(x,y) = K y   (when y is different from x)
A(x,MN) = S (A(x,M)) (A(x,N))

Lambda-terms are closed when they have no "undefined variables". For example, the lambda-term "λx.λy.x" is closed, while "λy.x" is not. For more details on lambda-calculus, please see the literature and/or search the WWW (though these are not required for participating in this contest).]

Submission

Player programs must be submitted via:

https://spreadsheets.google.com/spreadsheet/viewform?formkey=dDRFOXgxQ3g2M3I3X1VTV3FBREdULWc6MQ
(or http://goo.gl/FTP9F)

Each submission must be a single .tar.gz compressed archive file containing (at least) two executable files "./install" and "./run", and a "./src/" directory (all without the quotes). The "install" file will be executed just once with no human interaction after the submission is expanded into the home directory of a unique general user. The execution environment has been announced at:

http://www.icfpcontest.org/2011/05/judges-machine-and-system-environment.html
(or http://goo.gl/2DBgy)

Please try to minimize the installation procedure: for example, avoid compiling your program from source code and include a precompiled binary as far as possible.

The "run" file will be executed for each match with a command-line argument "0" (without the quotes, ditto below) or "1", meaning whether it is executed as player 0 or player 1 of the match. For each turn of the player, "run" must write to standard output either (as determined by the player) of:

  • the digit "1" (meaning left application) and the line feed character '\n', followed by a case-sensitive string (meaning a card name) and '\n', again followed by a decimal integer from 0 to 255 (meaning a slot number) and '\n', and
  • the digit "2" (meaning right application) and the line feed character '\n', followed by a slot number and '\n', again followed by a card name and '\n'.

For each turn of the opponent, "run" must read from standard input either (as determined by the opponent) of:

  • the digit "1" (meaning left application) and the line feed character '\n', followed by a card name and '\n', again followed by a slot number and '\n', and
  • the digit "2" (meaning right application) and the line feed character '\n', followed by a slot number and '\n', again followed by a card name and '\n'.

The input and output must be in ASCII code. Each output by the player must be performed within 60 seconds (in real time) after the last output from the opponent. In addition, for each match, the total CPU usage, the maximum resident set size, and the maximum disk usage of "run" must not exceed 10000 seconds, 512 MB, and 1 GB, respectively, including any descendant (child, grandchild, grand-grandchild, etc.) processes. Any files created by "run" during a match must be deleted by itself before or when it terminates. "run" must not communicate any information across matches. Submissions that are found to violate any of the above requirements would lose the match and/or be disqualified from the entire contest. [Remark: The judges also reserve the right to disqualify any submissions that we find are cheating and/or harmful; use common sense.]

[Remark: Although not mandatory, it is strongly urged that submissions are tested in advance on Debian squeeze for amd64 (which supports x86-64 including Intel 64). Details of the environment are described in:

http://www.icfpcontest.org/2011/05/judges-machine-and-system-environment.html
(or http://goo.gl/2DBgy)

If necessary, use some virtual machine software such as VirtualBox and VMware Player. See our unofficial information:

http://www.icfpcontest.org/2011/06/how-to-prepare-environment-for-testing.html
(or http://goo.gl/whrS7)

In case you can only produce 32-bit executables, please remember:

  • Statically linked 32-bit Linux executables should run with no problem on the 64-bit environment;
  • Dynamically linked 32-bit libraries should also be fine as long as they are included in the ia32-libs package of Debian; and
  • other dynamically linked 32-bit libraries (and any other files required at runtime but unavailable on Debian squeeze for amd64) can be included in the submission itself.

]

The "src" directory must contain all the source code written for the submission. It does not have to include publicly available common software such as standard compilers and libraries. (The judges will not try to compile the source code, though winning programs will be examined more carefully.)

[Remark: Here is a (very) simple example of the "install" script:

#!/bin/sh
exit 0

with an example of the "run" program:

#!/bin/sh
opp()
{
  read lr
  case $lr in
    1) read card; read slot;;
    2) read slot; read card;;
  esac
}
if [ $1 = "1" ]; then
  opp
fi
while [ true ]; do
  echo "1"
  echo "I"
  echo "0"
  opp
done

]

Additionally, the archive may include an optional "./README" file, explaining the most interesting aspect(s) of the submission. This file will be used for determining the discretionary judges' prize.

Evaluation

Submitted programs are evaluated in two rounds. In the first round, each program plays as player 0 against n randomly chosen distinct programs as player 1. The number n will be at least 15 and determined by the judges considering the number of submissions and the actual time required for the matches. In each match, 6 (resp. 2) points are given to player 0 if it wins against player 1 within (resp. after) 100000 turns, and 1 point is given to player 0 in case of a tie (in this round, no point is given to player 1). The programs are ranked by the total points.

In the second round, the best 30 (or more, in case of ties) programs in the first round play against each of the other 29 (or more) programs to determine the 1st and 2nd places. In each match, 6 (resp. 2) points are given to a program if it wins against another program within (resp. after) 100000 turns, and 1 point is given to both programs in case of a tie. The programs are ranked by the total points from the second round (points from the first round are not included). [Remark: In case of ties for the 1st and/or 2nd places, the prize(s) will be divided equally. Details of this division are at the discretion of the judges.]

Judges' prize will be determined discretionarily on the basis of the optional README files and the submissions themselves.

Enjoy!

Questions about this task description can be posted as (moderated) comments to this post or e-mailed to icfpc2011-blogger AT kb.ecei.tohoku.ac.jp. Please use the latter if your question includes spoiler information.

[Remark: Use #icfpc2011 for Twitter and #icfp-contest (on chat.freenode.net) for IRC, though the contest organizers will not always watch them (so do not ask questions there). Do not publish or communicate spoilers anywhere, either (again, please use common sense). Thank you!]

[Remark: An unofficial duel server is available at http://kokako.kb.ecei.tohoku.ac.jp/. It is unofficial in the sense that uploading there does not mean a submission for the official contest, that details may differ from the official rules, and that its availability and correctness are not guaranteed (though reports of bugs and problems are welcome - please comment to this post or e-mail to icfpc2011-blogger AT kb.ecei.tohoku.ac.jp).]

139 comments:

  1. Cool problem! Thanks for coming up with it.

    ReplyDelete
  2. It says that a turn ends when an error is raised, and errors can be raised by automatic applications of slots with -1 vitality, but do not affect other automatic applications. Do these errors end the turn — i.e., can you prevent your opponent from playing by zombieing their slots with crashers?

    ReplyDelete
  3. No, as the task description says, the automatic applications of zombies happen _before_ a turn and do not prevent it.

    ReplyDelete
  4. It's not clear how we can see what the opponent has performed during their turn.

    ReplyDelete
  5. I know it says that card images are just for amusement, but it looks like the image for "put" card does not correspond to description.
    The image says the put x y = return y, but description says that put takes only one argument (unused) and returns identity function, i.e. image should say put x = return I

    ReplyDelete
  6. > I know it says that card images are just for amusement, but it looks like the image for "put" card does not correspond to description.

    You are right. The task description is correct. We'll fix the image to avoid confusion. Thanks!

    ReplyDelete
  7. I'd like to clarify `attack i j n' and `help i j n'. Each of these involves changes to some slot `i' and another slot `j'. If `i' is not a valid slot number, do the changes to slot `j' still take place before the exception is raised? If `j' is not a valid slot number, do the changes to slot `i' still take place before the exception is raised? Thanks.

    ReplyDelete
  8. Please read the task description carefully. The order of such effects
    are exactly described.

    ReplyDelete
  9. It would help a lot if unofficial server could show logs. Currently my submission loses by invalid output at turn 4000+ and I can't figure out what's wrong, because all it does now is spamming the same command. :(

    ReplyDelete
  10. We cannot publish the original logs for the obvious reason (they
    reveal the strategy of other players). We will think about better
    output perhaps.

    ReplyDelete
  11. For attack and help cards, one of the steps is "do nothing the the slot is dead". Does this refer the the proponent's slot i, the opponent's slot (255 - j), or the slot currently being either right or left applied?

    ReplyDelete
  12. It means the opponent's (255-j)th slot in "attack" and the proponent's jth slot in "help".

    ReplyDelete
  13. What is a directory where "run" file is executed?

    ReplyDelete
  14. Question regarding automatic applications. Suppose "attack" is applied to some value, yielding another function. If this function is applied during an automatic application, will the behavior be the modified one or the original one?

    ReplyDelete
  15. > Question regarding automatic applications. Suppose "attack" is applied to some value, yielding another function. If this function is applied during an automatic application, will the behavior be the modified one or the original one?

    The former, that is, the behavior changes.

    ReplyDelete
  16. > What is a directory where "run" file is executed?

    Your home directory, that is, the home directory of the unique general user assigned to your submission, in which the .tar.gz archive is expanded.

    ReplyDelete
  17. Is there any validation tool to ensure that submission is correct (in context of its format).

    ReplyDelete
  18. If slot k has vitality of (-1) before a turn, and as a result of an automatic application of slot j (<k) k's vitality is no longer (-1), does an automatic application still happen on k?

    ReplyDelete
  19. > Is there any validation tool to ensure that submission is correct (in context of its format).

    No, please use your own testing environment:

    http://www.icfpcontest.org/2011/06/how-to-prepare-environment-for-testing.html

    and perhaps the _unofficial_ duel server:

    http://kokako.kb.ecei.tohoku.ac.jp/

    ReplyDelete
  20. If I steal a revive card from the opponent using copy, will it revive my slots or him/her when applied?

    ReplyDelete
  21. > If I steal a revive card from the opponent using copy, will it revive my slots or him/her when applied?

    The former. "copy" only steals a value (not the effect), and the effects of cards are fixed (e.g. "revive" always works on the proponent, i.e. the player who applies it).

    ReplyDelete
  22. > If slot k has vitality of (-1) before a turn, and as a result of an automatic application of slot j (<k) k's vitality is no longer (-1), does an automatic application still happen on k?

    No, the vitality is checked before _each_ automatic application. The task description is now clarified (Update on 07:05 June 17 Friday UTC). Thanks!

    ReplyDelete
  23. Please correct me if I'm wrong.
    I suppose the following scenario describes zombie functioning.

    ...
    Player0 raises a zombie in player1's 13th slot.
    At the beginning of the next player1's turn this zombie is applied with proponent=player1 and opponent=player0, so when zombie executes 'inc', player1 suffers.

    ReplyDelete
  24. > Please correct me if I'm wrong. I suppose the following scenario describes zombie functioning. ... Player0 raises a zombie in player1's 13th slot. At the beginning of the next player1's turn this zombie is applied with proponent=player1 and opponent=player0, so when zombie executes 'inc', player1 suffers.

    Yes.

    ReplyDelete
  25. About attack: can one suicide using it (by having n equal to v) or not? In other word, is the "greater" in the text "strictly greater" or "greater or equal"?

    ReplyDelete
  26. > About attack: can one suicide using it (by having n equal to v) or not? In other word, is the "greater" in the text "strictly greater" or "greater or equal"?

    The "greater" means "strictly greater".

    ReplyDelete
  27. BTW, will there be a "chmod +x install run" after unpacking the submission from archive?
    Unofficial server seems not to do it (which makes a lot of pain for Windows users)

    ReplyDelete
  28. Hi, I'm the one who complained about "put" car image being wrong, but now I realized that "put x y = y" and "put x = return I" are equivalent definitions for put function. But the latter is probably less confusing and definitely more explicit. Thanks for listening!

    ReplyDelete
  29. > BTW, will there be a "chmod +x install run" after unpacking the submission from archive?

    Yes.

    ReplyDelete
  30. >> BTW, will there be a "chmod +x install run" after unpacking the submission from archive?
    >
    >Yes.

    So "unofficial" server does that too, right?

    ReplyDelete
  31. It seems (by experimenting with the interactive LTG executable), that the second argument of K is not evaluated (side effects do not appear). Is this correct?

    ReplyDelete
  32. > So "unofficial" server does that too, right?

    Yes. (The line feed code problem, perhaps? Please use LF, not CR+LF.)

    ReplyDelete
  33. No, the problem is "player 1 played an invalid hand" / "player 1 loses by invalid output at turn xxxx".
    Which is very weird to happen as all the program does is reading 3 lines and spamming the same three lines all over again.

    Can it be due to 60 seconds limit?

    ReplyDelete
  34. > It seems (by experimenting with the interactive LTG executable), that the second argument of K is not evaluated (side effects do not appear). Is this correct?

    The application is performed as explained in the task description, namely, call by value.

    ReplyDelete
  35. > No, the problem is "player 1 played an invalid hand" / "player 1 loses by invalid output at turn xxxx".
    Which is very weird to happen as all the program does is reading 3 lines and spamming the same three lines all over again.

    Resource limits in the unofficial duel server turned out to be too restrictive than we expected. We have changed the limiting mechanism now, so please resubmit
    the solution if the "invalid hand" seems to be a false alert. Thank you for cooperation.

    ReplyDelete
  36. I am currently installing Deibian in VirtualBox, but am stuck at "Install the GRUB boot loader on a hard disk."

    The message box below then tells me that Debian detects that it is the only operating system on the computer (It's not, I'm VMing in windows), and then asks me if Debian should "Install the GRUB bot loader to the master boot record?" with a Yes/No option.

    Should I allow Debian to do this?

    ReplyDelete
  37. >Card "copy" is a function that takes an argument i, and returns the value of the field of the ith slot of the opponent. It raises an error if i is not a valid slot number. Note that the slot is ith, not (255-i)th.

    If the ith slot of the opponent is dead, does an error raise?

    ReplyDelete
  38. A few questions.

    About zombies and attack: do I read correctly that the opponent who is running the zombie a slot decreased, and the proponent has a slot 9/10 increased?

    About attack and health: if 'j' does not denote a correct slot number, do I read correctly that 'i' is decreased nonetheless?

    Thanks,

    Alan

    ReplyDelete
  39. > If the ith slot of the opponent is dead, does an error raise?

    No, it raises an error only if i is not a valid slot number.

    ReplyDelete
  40. on submission page there is "URL of your submission".
    does this mean that i will have to upload the tar.gz file somewhere on www and give link to that location here?

    ReplyDelete
  41. > About zombies and attack: do I read correctly that the opponent who is running the zombie a slot decreased, and the proponent has a slot 9/10 increased?

    Yes.

    > About attack and health: if 'j' does not denote a correct slot number, do I read correctly that 'i' is decreased nonetheless?

    Yes.

    ReplyDelete
  42. I have a question about the dynamically bound cards inc, dec, attack, help which change their behaviour during the pre-turn run.

    When one of these is partially applied during a turn (in context A), say "attack 0" and later during the turn (in context B) the remaining arguments are supplied, which behaviour should I see?

    The card description suggests I will see behaviour depending on context A but LTG shows partially applied function by name (ie. "attack 0" rather than "\j -> \n -> ...") suggesting the name might not be bound until context B.

    ReplyDelete
  43. > on submission page there is "URL of your submission".
    > does this mean that i will have to upload the tar.gz file somewhere on www and give link to that location here?

    Yes. (Please use any free service if you do not have a server.)

    ReplyDelete
  44. > I have a question about the dynamically bound cards inc, dec, attack, help which change their behaviour during the pre-turn run.
    >
    > When one of these is partially applied during a turn (in context A), say "attack 0" and later during the turn (in context B) the remaining arguments are supplied, which behaviour should I see?
    >
    > The card description suggests I will see behaviour depending on context A but LTG shows partially applied function by name (ie. "attack 0" rather than "\j -> \n -> ...") suggesting the name might not be bound until context B.

    It is the latter, that is, what you describe "bound under context B" (but please note that the task description does not use such a notion as binding of cards or their names).

    ReplyDelete
  45. Could it be possible to have a little bit of information as to why a run is "invalid" on the test server? It could be "Broken Pipe" (the program quit), "Timeout" (took too long), or "Parsing" (could not parse the output).

    I know the test server is unofficial, but it's hard to debug when the only information is "does not work" (ask a sysadmin ;-)

    ReplyDelete
  46. We have just noticed that our runtime requires X to be running; since some errors are reported with dialogs. Will the test server have an X server available?

    ReplyDelete
  47. What if exactly 1000th application has side effect? Will evaluation be aborted before or after that application?

    ReplyDelete
  48. > What if exactly 1000th application has side effect? Will evaluation be aborted before or after that application?

    Not before. Please read the Rules carefully.

    ReplyDelete
  49. I second the question: what is "invalid"
    (on the unofficial server)?

    The word "invalid" does not appear in the spec,
    and the ltg.linux executable don't seem to emit it.

    The spec says that a turn ends when an error is
    raised, but this does not end the match?

    ReplyDelete
  50. Is there any chance you could add a rankings page to the unofficial duel server showing each team's current submission ranked according to the percentage of duels won? It would be nice to see at a glance how we are doing compared to other teams.

    ReplyDelete
  51. > I second the question: what is "invalid"
    (on the unofficial server)?

    A duel result is counted as #Invalid when either of them happens:
    1) Running 'install' failed.
    2) Running 'run' failed.
    3) The player's output did not conform to the format specified in the "Submission" section.
    4) The player exceeds time limit or memory limit.

    If a submission performed something invalid 3 times, that submission will not be used in the unofficial duel server anymore.

    > The spec says that a turn ends when an error is
    raised, but this does not end the match?

    It does not end the match.

    ReplyDelete
  52. Another #invalid:
    0) Submission is not .tar.gz

    (I didn't carefully read the spec and submitted a zip file the first time.)

    ReplyDelete
  53. Per the response to Vlad's comment above, I see where the spec says "... exceeds 1000", but it is still not perfectly clear whether the side effects of the 1001st application (if any) persist.
    Do they?

    Technically, I recognize that to "exceed 1000" applications, the 1001st application must have taken place. But, since function application seems to be an evaluation process, which may sometimes terminate abruptly, it is unclear whether the application has taken place at the beginning of evaluation (before side-effects), or only when evaluation is complete (after side-effects).

    ReplyDelete
  54. > Another #invalid:
    0) Submission is not .tar.gz

    Right. Thank you for pointing it out.

    ReplyDelete
  55. > Could it be possible to have a little bit of information as to why a run is "invalid" on the test server? It could be "Broken Pipe" (the program quit), "Timeout" (took too long), or "Parsing" (could not parse the output).

    We are now considering to add such discrimination in the unofficial server log (but it could take time).

    ReplyDelete
  56. > We have just noticed that our runtime requires X to be running; since some errors are reported with dialogs. Will the test server have an X server available?

    We have installed xfvb package on the unofficial duel server.
    We will add it to the official judge machine image, too.

    ReplyDelete
  57. > Per the response to Vlad's comment above, I see where the spec says "... exceeds 1000", but it is still not perfectly clear whether the side effects of the 1001st application (if any) persist.
    > Do they?

    When the 1000th application is triggered, its accompanying side-effects
    always take place. When the 1001st application is triggered, the
    turn ends immediately without any side-effects accompanying the 1001st application.

    ReplyDelete
  58. If I start with (K zero) in a slot and in the current turn I apply that slot to a zero card, how many function applications are counted in the current turn?

    ReplyDelete
  59. Zere is probably one problem with zero card. Correct me if I'm wrong.

    Accordingly to spec zero is always returning 0 number. However if you will call zero(I) or any other function as parameter - Native. Exception is raised. Is it expected?

    It looks like zero expect to see only numbers. Why?

    Thanks in advance

    ReplyDelete
  60. If we start with ((S S) I) in a slot and in the current turn we apply that slot to a "I" card, how many function applications are counted in the current turn?

    ReplyDelete
  61. 0 is just a constant, not a function.

    ReplyDelete
  62. > If we start with ((S S) I) in a slot and in the current turn we apply that slot to a "I" card, how many function applications are counted in the current turn?

    Four.

    ReplyDelete
  63. Just to clarify can I submit updates of my package to official server after first submission?

    ReplyDelete
  64. > Just to clarify can I submit updates of my package to official server after first submission?

    Yes, you can update the submission by using the same password.

    ReplyDelete
  65. Again about zero:
    does evaluating zero cost one application?
    e.g. does 'succ(zero)' cost 1 or 2 applications?

    ReplyDelete
  66. > Again about zero:
    does evaluating zero cost one application?
    e.g. does 'succ(zero)' cost 1 or 2 applications?

    No. The cost for 'succ(zero)' is 1.

    ReplyDelete
  67. > Is there any chance you could add a rankings page to the unofficial duel server showing each team's current submission ranked according to the percentage of duels won? It would be nice to see at a glance how we are doing compared to other teams.

    We have set up a ranking page for the unofficial server: http://kokako.kb.ecei.tohoku.ac.jp/leaderBoard.
    Thanks for your suggestion.

    ReplyDelete
  68. About the invalid thing: if the opponent outputs an invalid string, will it be passed to the proponent before it is checked?

    To give a concrete example: when we read what the opponent is doing, we fail (and exit the program) if it is not a correct output. Could it happen that this happens before the opponent's play is checked (thus the fact we quit is detected as invalid play)?

    ReplyDelete
  69. > About the invalid thing: if the opponent outputs an invalid string, will it be passed to the proponent before it is checked?

    No. The opponent outputs are first checked before they are passed to the proponent.

    ReplyDelete
  70. What is the output of the test server when the program takes too long? Is it also "player ... loses by invalid output at turn ..."?

    ReplyDelete
  71. In the submission page (http://goo.gl/FTP9F)
    We have to
    >Copy-and-paste the first column of the output of "sha512sum" on your tar.gz file

    I want to make it clear about what the "first column" is.
    For example, if I use gpg to the file "submit"

    >gpg --print-md sha512 submit
    submit: CF83E135 7EEFB8BD F1542850 D66D8007 D620E405 0B5715DC 83F4A921 D36CE9CE
    47D0D13C 5D85F2B0 FF8318D2 877EEC2F 63B931BD 47417A81 A538327A F927DA3E

    Is the "first column" CF83E135 in this case?

    ReplyDelete
  72. > I want to make it clear about what the "first column" is.

    We have elaborated the instructions on the submission page:

    > Copy-and-paste the first column (128 hex digits in lower case with no spaces) of the output of the "sha512sum" command (on Debian squeeze or any compatible environment) on your tar.gz file.

    ReplyDelete
  73. > What is the output of the test server when the program takes too long? Is it also "player ... loses by invalid output at turn ..."?

    Currently yes. We will update the error message but it may take time.

    ReplyDelete
  74. A question about the description of the zombie card: the description says that zombie raises an error if i is not a valid slot number. Isn't it (255-i)? According to the description, it seems that zombie sets the vitality of the i-th slot to -1 although the image implies that the vitality of the (255-i)th slot should be set to -1.

    ReplyDelete
  75. > A question about the description of the zombie card: the description says that zombie raises an error if i is not a valid slot number. Isn't it (255-i)? According to the description, it seems that zombie sets the vitality of the i-th slot to -1 although the image implies that the vitality of the (255-i)th slot should be set to -1.

    No, "the slot" in "zombie" refers to "the (255-i)th slot". Note that i is an integer, not a slot. Note also that i is a valid slot number if and only if 255-i is.

    ReplyDelete
  76. Question about Zombie.
    During automatic apply we get:
    X(I) or I(X) ?
    where X is the function we set with Zombie card and I is Identity function.

    ReplyDelete
  77. So when "run" is executed with parameter "0" it should write three lines, then wait for three lines, then write three more lines and so on, while when it is executed with parameter "1" it should wait for three lines first.

    Am I correct?

    Should I expect the same behaviour from, say, ltg.linux32? (i.e. are any other fancy output in ltg.linux32 written to stderr?)

    ReplyDelete
  78. could you please tell me the path of python2.6 (opposed to python3) interpreter in the runtime environment?

    ReplyDelete
  79. > So when "run" is executed with parameter "0" it should write three lines, then wait for three lines, then write three more lines and so on, while when it is executed with parameter "1" it should wait for three lines first.
    >
    > Am I correct?

    Correct.

    > Should I expect the same behaviour from, say, ltg.linux32? (i.e. are any other fancy output in ltg.linux32 written to stderr?)

    Yes.

    ReplyDelete
  80. > Question about Zombie.
    > During automatic apply we get:
    > X(I) or I(X) ?
    > where X is the function we set with Zombie card and I is Identity function.

    Please read the task description carefully: "applying f to x" means "f(x)", not "x(f)".

    ReplyDelete
  81. > could you please tell me the path of python2.6 (opposed to python3) interpreter in the runtime environment?

    /usr/bin/python2.6

    /usr/bin/python is also python 2.6.6

    ReplyDelete
  82. I'm getting a "500 Internal Server Error" when trying to upload my package to the unofficial duel server.

    ReplyDelete
  83. Duel server is giving internal error when trying to upload a bot.

    ReplyDelete
  84. > I'm getting a "500 Internal Server Error" when trying to upload my package to the unofficial duel server.
    > Duel server is giving internal error when trying to upload a bot.

    Thank you for your reports. We fixed the problem.

    ReplyDelete
  85. What does the "Sorry, there is a misjudgement" result message mean ?

    ReplyDelete
  86. > > About attack: can one suicide using it (by having n equal to v) or not? In other word, is the "greater" in the text "strictly greater" or "greater or equal"?

    > The "greater" means "strictly greater".

    The ltg simulator does not implement this: it is possible to suicide doing an "attack" or a "help".

    ReplyDelete
  87. I'm getting "player 1 loses by invalid output at turn 1" on the duel server every time. My submission works just fine when running with the 32bit ltg on a debian virtual machine.

    Latest submissio ID is 919.

    Any help would be greatly appreciated

    ReplyDelete
  88. > What does the "Sorry, there is a misjudgement" result message mean ?

    We had a server problem at that time. Please just ignore the entry. It shouldn't affect the number of wins/losses.

    ReplyDelete
  89. > the digit "1" (meaning left application) and the line feed character '\n', followed by a card name and '\n', again followed by a slot number and '\n', and

    It's not clear if card name is case sensitive or not. For output data it is clarified, for input it is not.

    ReplyDelete
  90. > > > About attack: can one suicide using it (by having n equal to v) or not? In other word, is the "greater" in the text "strictly greater" or "greater or equal"?

    > > The "greater" means "strictly greater".

    > The ltg simulator does not implement this: it is possible to suicide doing an "attack" or a "help".

    Please read the task description more carefully before asking questions. Thank you for your consideration.

    ReplyDelete
  91. Team Lambda Lambda Kurgan!June 19, 2011 at 2:46 AM

    Hi,
    >>Any files created by "run" during a match must be deleted by itself before or when it terminates.

    Does this include interpreter tokenized files such as *.pyc created by Python ?

    - Team Lambda Lambda Kurgan! - there can only be one!
    P.S. Thank you for a wonderful contest!

    ReplyDelete
  92. > It's not clear if card name is case sensitive or not. For output data it is clarified, for input it is not.

    Card names are case-sensitive both for input and output. In section "Submission", "a card name" is "a case-sensitive string" as defined at its first occurrence.

    ReplyDelete
  93. > I'm getting "player 1 loses by invalid output at turn 1" on the duel server every time. My submission works just fine when running with the 32bit ltg on a debian virtual machine.
    >
    > Latest submissio ID is 919.
    >
    > Any help would be greatly appreciated

    Please read the paragraph "In case you can only produce 32-bit executables [...]" in the task description.

    ReplyDelete
  94. > >>Any files created by "run" during a match must be deleted by itself before or when it terminates.

    > Does this include interpreter tokenized files such as *.pyc created by Python ?

    In general, "./run" should delete any temporary files it creates during a match. If you would like to share any files (such as precompiled bytecode) across matches, consider including them in the submission itself (or generating them during "./install").

    ReplyDelete
  95. The Invisible ImpJune 19, 2011 at 5:10 AM

    Thank you for creating the leader board!

    I'm a little confused by its output, though. How does it pick which 30 duels to count? How frequently does it update?

    I would have expected it to use the most recent 30, updating whenever a duel completed... but that's clearly not the case, as my most recent submission has won all its matches since submission (OK, that's only 2 ;-) ), but I'm still listed as having 0 wins.

    My team name is The Invisible Imp, if that helps at all...

    ReplyDelete
  96. > I'm a little confused by its output, though. How does it pick which 30 duels to count? How frequently does it update?

    As noted at the top of the leader board, the latest 30 valid duels in the last 12 hours of each team are considered. The ranking is updated every 10 minutes. These number may change, though.

    ReplyDelete
  97. Is it acceptable for ./run to attempt to play indefinitely, relying on the game server to kill it, or must ./run terminate itself when a game is won or tied?

    ReplyDelete
  98. The text for the "S" card in the specification says that it applies f to x before applying g to x.

    However, the ltg executable appears to do the reverse.

    Can you confirm that the specification is accurate about this?

    ReplyDelete
  99. > Can you confirm that the specification is accurate about this?

    The specification is correct as it is and we are sure that ltg conforms to it.

    ReplyDelete
  100. for help command, can i == j? If yes, does it the count addition after the subtraction?

    ReplyDelete
  101. also, can n for attack and help be 0?

    ReplyDelete
  102. I'm stuck with the submission to the unofficial server. I converted the run example script you provided into python and submitted it with your install script. I get the result/log:

    submission does not conform to the specification/invalid output at turn 1

    submission ID is 1221... can you give me a hint?

    ReplyDelete
  103. > for help command, can i == j? If yes, does it the count addition after the subtraction?

    Yes and yes.

    ReplyDelete
  104. > also, can n for attack and help be 0?

    Yes.

    ReplyDelete
  105. > Is it acceptable for ./run to attempt to play indefinitely, relying on the game server to kill it?

    Yes.

    ReplyDelete
  106. > submission does not conform to the specification/invalid output at turn 1

    Did you flush stuout?

    ReplyDelete
  107. "submission does not conform to the specification" message is the result of Native.Error ?
    For example if apply card to dead slot it would be just reset to Identity function or player will lose with "submission does not conform to the specification" message ?

    P.S. Anyway, could you replace "submission does not conform to the specification" message with something a bit more informative ?

    ReplyDelete
  108. > "submission does not conform to the specification" message is the result of
    > Native.Error ?

    No.

    > For example if apply card to dead slot it would be just reset to Identity
    > function or player will lose with "submission does not conform to the
    > specification" message ?

    Former.

    Currently "submission does not conform to the specification" means
    either of the following:
    1) your program's output does not follow the format written in the "Submission" section, or your program does not output anything (please flush stdout!).
    2) your program used 60 seconds in one turn.

    > P.S. Anyway, could you replace "submission does not conform to the
    > specification" message with something a bit more informative ?

    We are trying to update the error message to be more informative,
    however please understand some unrecognizable error may still occur.

    ReplyDelete
  109. The Invisible ImpJune 19, 2011 at 11:14 PM

    If a submission emits text to stderr, will that be ignored, or will that cause a "does not conform to the specification" error?

    ReplyDelete
  110. Does Ruby environment works on unofficial server?
    My submission works fine in ltg match mode, but when uploaded it fails with "submission does not conform to the specification" error

    ReplyDelete
  111. > If a submission emits text to stderr, will that be ignored, or will that cause a "does not conform to the specification" error?

    It will be ignored.

    ReplyDelete
  112. > Does Ruby environment works on unofficial server? My submission works fine in ltg match mode, but when uploaded it fails with "submission does not conform to the specification" error

    Ruby should work on the unofficial server. If you provide your team name or submission id to us (by email, perhaps?), we will investigate the reason.

    ReplyDelete
  113. Can I resubmit a new solution after submitting one via the Google Docs form?

    ReplyDelete
  114. Will unofficial duel server work after contest end? It is an interesting game, so we can compete even after contest.

    ReplyDelete
  115. > Can I resubmit a new solution after submitting one via the Google Docs form?

    Yes, you can.

    ReplyDelete
  116. On unofficial server you don't get win if your opponent exceeds time or memory limit. For example duel ID = 21593.
    Test it on official server please:)

    ReplyDelete
  117. What will you do if after n games of first round where would be several tied teams for 30th place?

    ReplyDelete
  118. > On unofficial server you don't get win if your opponent exceeds time or memory limit.

    You are right -- we'll make sure that you win in such a case in official matches.

    ReplyDelete
  119. > What will you do if after n games of first round where would be several tied teams for 30th place?

    Please read the task description (the section on Evaluation).

    ReplyDelete
  120. Shouln't that text on the bottom of the image of Card "get x" read "Memorize ..." instead of "Memoize ..." ?
    Or could you provide me with the meaning of the word memoize? My dictionary could not find it.
    I like your Card images very much!

    ReplyDelete
  121. The unofficial server says that the leaderboard has been "Frozen at 15:17:26 GMT, June 20", however the stated time has not yet past. (I write this at Sun Jun 19 17:26:49 UTC 2011.)

    ReplyDelete
  122. Are there any notifications about done submission except
    "Thank you for your participation to ICFP Programming Contest 2011!" on the web form. Email?

    ReplyDelete
  123. > Shouln't that text on the bottom of the image of Card "get x" read "Memorize ..." instead of "Memoize ..." ?
    > Or could you provide me with the meaning of the word memoize? My dictionary could not find it.

    http://www.google.com/search?q=memoize

    ReplyDelete
  124. > The unofficial server says that the leaderboard has been "Frozen at 15:17:26 GMT, June 20", however the stated time has not yet past. (I write this at Sun Jun 19

    Fixed. Thanks!

    ReplyDelete
  125. > Are there any notifications about done submission except
    > "Thank you for your participation to ICFP Programming Contest 2011!" on the web form. Email?

    No, but we have elaborated the message after submission to:

    > Thank you for your participation to ICFP Programming Contest 2011! We will download your submission within a week. (In case of problems, we may ask questions to the e-mail address you registered.)

    ReplyDelete
  126. Does slot function is reset to I when slot becomes dead or revived ?

    ReplyDelete
  127. Does dead slot field will be reset when your perform left or right application to dead slot ?

    ReplyDelete
  128. > Does slot function is reset to I when slot becomes dead or revived ?

    No. See the task description.

    > Does dead slot field will be reset when your perform left or right application to dead slot ?

    Yes. Please read the task description carefully.

    ReplyDelete
  129. If we implemented two bots, is it allowed to submit more than one solution per team?

    ReplyDelete
  130. > If we implemented two bots, is it allowed to submit more than one solution per team?

    No. Please choose your champion bot!

    ReplyDelete
  131. What will happen if someone other will re-submit solution under our name? As we've investigate, there is no handling of such situations.

    ReplyDelete
  132. Dear organizers,

    Will you keep the server alive for a few more days or weeks, so we may try to hack this SKI completely? Or would you mind to publish source code and submitted bots somewhere somewhen?

    ReplyDelete
  133. > What will happen if someone other will re-submit solution under our name? As we've investigate, there is no handling of such situations.

    We basically distinguish submissions from different teams in terms of passwords.
    In suspicious cases, we will handle the situation individually by contacting the teams by emails.

    ReplyDelete
  134. Could you give some details about when to expect the final and preliminary results and what information will be in preliminary results?

    ReplyDelete
  135. While I am not going to submit my final product I want to thank the organizers for providing a fun, educational, interesting diversion for my free time this weekend.

    Thanks,
    Graham (aka Slow)

    ReplyDelete
  136. The Invisible ImpJune 20, 2011 at 1:36 PM

    I also would like to thank the organizers of this contest. It was definitely challenging! Further, I'd like to thank them for leaving the duel server running; it's been nice to see the data slowly accumulate on the win/loss rate for my final submission, even past the end of the contest.

    I look forward to seeing the announcement of the final rankings of the official submissions... it's just sad that I'll have to wait for three months to see it.

    Could you please announce when you've completed fetching all the submission tarballs? I'd be a nice confirmation.

    ReplyDelete
  137. Nice contest :)
    Even for non-functional programmers :)

    Good Job!

    There's will be nice if you'll provide some information about submissions fetching.

    I.e. could you please, notify users when their's submission will be fetched by email, or on the site (Marking +/- in teamlist or something)

    Thank you again :)

    ReplyDelete