Tagging
It is possible to track the identity of a particular piece.
Each piece in the initial position of the game or study is assigned a unique nonnegative integer, called its mark.
Promotion does not change the mark of a piece.
A tag is a named mark. We say that a tag is bound to the value of a mark.
A tag represents the particular piece whose mark equals the mark of the tag. The following keywords manipulate the mark:
:forany in the match list
A tag is defined by the :forany
keyword in a match list.
The :forany
keyword takes two parameters: the alphanumeric name of a tag and a piece designator.
The match filter is then conceptually run once in which each tag is bound to each possible mark, such that two conditions hold:
- No two tags may be bound to the same mark.
- A tag may only be bound to a mark that represents a piece that either matches, or eventually matches, the piece designator associated with that tag.
Square information associated with that piece designator is ignored.
At most two tags may be contained inside a match list.
Tags may not be used in a matchlist if a :relation occurs in a position list inside that match list.
:tagmatch in the position list
A :tagmatch
keyword in a position list takes two parameters, the name of a tag and a piece designator.
The position list match fails on the current position if the piece and square corresponding to the mark of the tag does not match the given piece designator.
For example, assuming there is a tag named "foo", this position will only match if "foo" is a white pawn:
(position :tagmatch foo P)
Piece designators and tags
A piece designator can have a piecetype designator that is the name of a tag preceded by the character '$'.
Such a piecetype represents the particular piece represented by the named tag.
Because tag names are alphanumeric, square designators in such a piece designator must be enclosed in brackets.
Example:
(position $foo[a1,d4])
The piece represented by the tag named "foo" must be on a1
or d4
.
Longer example
Consider the problem of recognizing all studies or games in which the same rook visits all four corners of the board.
This is shown in the file CQL Rook visits each corner, and is reproduced here:
; Find all games in which the same rook visits all four corners of the board. (match :pgn heijden.pgn :output out.pgn :forany rook [rR] ; loop over the possible rooks (position $rook[a1]) (position $rook[h1]) (position $rook[h8]) (position $rook[a8]) )
Here, the :forany
keyword in the match list introduces a tag named rook
which is constrained to range over white or black rooks, or over pawns that eventually promote to one of those.
The first position list matches games in which there is a position such that the tag rook
represents a rook on a1.
The next position list matches games in which there is a postiion such that the tag represents a rook on h1.
The tag represents the same mark, that is, the same piece, between the position lists.
The :forany
keyword will sequentially bind the tag to each allowed mark.
Since a match list matches a game only if all its constituent position filters match it, the code does what we want here.
Other examples of tagging
The files
- Excelsior A pawn starts on its second rank and promotes.
- Double Excelsior Both a white and black pawn have excelsiors; the white pawn underpromotes.
- Many checks, The same piece delivers at least 30 checks in the game.
- Rook traverses a rectangle in this case, a 4x3 rectangle.
- Knight chases other Knight constantly trying to sacrifice itself.
- Knight visits at least twenty different squares.
present some other examples of tagging.