Converting old SGF files to FF[4]

Converting properties

This is a short list of properties that frequently occur in old SGF files and how they should be converted to FF[4].

Property:	M
Propvalue:	list of point
Function:	Simple board markup
Conversion:	M should be converted to MA if used on empty board points
		and to TR if used on stones.

Property:	L
Propvalue:	list of point
Function:	Label the given points with uppercase letters.
Conversion:	L should be converted to LB.
		Example: L[fg][es][jk] -> LB[fg:A][es:B][jk:C]

Property:	TE then BM within a node
Propvalue:	double
Function:	Move annotation
Conversion:	Should be converted to IT.

Property:	BM then TE within a node
Propvalue:	double
Function:	Move annotation
Conversion:	Should be converted to DO.


Property:	VW
Propvalue:	point, point
Function:	Restrict view (just like in FF[4]).
		VW wasn't an official property in FF[3] but was used by two
		applications (SGB & xgoban)
Conversion:	In FF[3] the two given points specified the upper
		left and the lower right corner. To convert it to FF[4] either
		make a compose type value out of it or write the uncompressed
		point list.
		Example: VW[ba][db] -> VW[ba:db]
			        or  -> VW[ba][ca][da][bb][cb][db]

Other conversion issues

Property identifiers

In FF[1]-FF[3] lowercase letters where allowed in property identifiers, but in FF[4] only uppercase letters are allowed. I.e. the application has to remove all lowercase letters from the file.

Example: Black[qd];thisisaWhitemove[kk]bbrWyryrerwLerreoi[10.3] --> B[qd];W[kk]WL[10.3]

Property types move & setup

The old fileformats didn't make this distinction, i.e. these files may have mixed setup and move-properties in one node.
In that case an application should split that node into two nodes: all setup, root and game-info properties and the node name ('N[]') into the first node, all other properties into the second. The node name should be in the first node because the node to be split might be the start of a variation.

Illegal & errornous files

There are a lot of corrupt files out there. If you want to write a robust application you should be able to deal with the following cases:

Line breaks & whitespaces within moves or values: just remove all whitespaces in property values except in text values.
Empty values e.g. 'DT[]N[]': remove those properties
Properties without values e.g. 'LB B[aa]': remove the property identifiers
Properties with too many values e.g. 'B[aa][bb]': this one is critical because it might be caused by a missing ']' e.g. 'B[aa LB[aa][bb]'
Empty variations e.g. '()': such a construct is illegal because a variation has to have at least one node - remove '()'
Files not starting with '(;' e.g. the ';' is missing - this one is tricky if you want to skip junk in front of the SGF file.
Faulty game-info values (e.g. RE, DT): the application should correct those values if possible, otherwise leave them as they are.

These are just some frequent errors that occur.
Have a look at the SGF Syntax Checker & Converter for more sophisticated errors and how to handle them.