You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Did not notice this last time I did major work back in Nov 2022.
Briefly it should accept queries as defined in version 6.0.4 of CQL. Home page is http://gadycosteff.com/cql/ but that has moved on to version 6.2 with a back-link as far as version 6.1.
At 6.2 the source code is available for download by default, no need to ask the authors, and because I see it build and run on OpenBSD, FreeBSD, and Debian (the sample Linux) it seems the command 'cql -parse -cql "statement"' run in a subprocess is a reasonable way to get a repr-like definition of the tokens and AST. (Binary downloads are available only for Windows and MAC.)
Plan is to see if chessql can be made to provide useful stuff with tokens supplied by the -parse option rather than 'tokens invented here', reverting to adding to existing stuff if forced. In practice that means 'will it do for chesstab'.
Commit ca33b23 parses the examples supplied with CQL 6.0.4 and CQL 6.1 but fails to parse the following examples supplied with CQL 6.2 in the examples and examples-ascii directories:
All but lacny are made to parse successfully by appending a newline to the example. However it is thought clearance-delayed would fail to parse except for incorrectly parsing the final '///' filter.
The 'path' and '///' filters, added at CQL 6.2, are terminated by '\n\n' and '\n' respectively assuming all filters following 'path' or '///' on the line are ok. Simple tests with the 'cql -parse' option suggest filter sequences such as 'path ... path ... /// ... path ... \n\n' or 'path ... /// ... \n ... path ... /// ... \n\n' or '/// "comment" \n' are allowed. '/// ... /// ... \n' is not allowed. Prior to CQL 6.2 '\n' was whitespace except for terminating a single line comment: now it is a terminator of filters where 'path' and '///' are involved.
Commit d33c52f parses the examples supplied with CQL 6.0.4, CQL 6.1 and CQL 6.2, but does not handle the '--' (utf8 '―') filter, introduced at CQL 6.2, correctly.
The initial question is why are six '--' filters found in the babson1 example when only two are in the text.
The second question is why adjusting the babson1 example so the '--' filters are not in the body of a function causes the correct number of '--' filters to be seen (two not six).
Activity
RogerMarsh commentedon Jul 24, 2024
Did not notice this last time I did major work back in Nov 2022.
Briefly it should accept queries as defined in version 6.0.4 of CQL. Home page is http://gadycosteff.com/cql/ but that has moved on to version 6.2 with a back-link as far as version 6.1.
At 6.2 the source code is available for download by default, no need to ask the authors, and because I see it build and run on OpenBSD, FreeBSD, and Debian (the sample Linux) it seems the command 'cql -parse -cql "statement"' run in a subprocess is a reasonable way to get a repr-like definition of the tokens and AST. (Binary downloads are available only for Windows and MAC.)
Version 6.1 http://www.gadycosteff.com/cql-6-1/ (I did not know this existed) mentions CQLi which has documentation in manual.pdf which looks worth having.
Plan is to see if chessql can be made to provide useful stuff with tokens supplied by the -parse option rather than 'tokens invented here', reverting to adding to existing stuff if forced. In practice that means 'will it do for chesstab'.
RogerMarsh commentedon Feb 1, 2025
Commit ca33b23 parses the examples supplied with CQL 6.0.4 and CQL 6.1 but fails to parse the following examples supplied with CQL 6.2 in the examples and examples-ascii directories:
babson3, checkcheckcheckstalemate, clearance-delayed, consecutive-checks-by-one-color, consecutive-checks, forced-moves-either-side, greek-gift, indian, lacny, likeinterference, rookleapcorners, and rundlaufe.
All but lacny are made to parse successfully by appending a newline to the example. However it is thought clearance-delayed would fail to parse except for incorrectly parsing the final '///' filter.
The 'path' and '///' filters, added at CQL 6.2, are terminated by '\n\n' and '\n' respectively assuming all filters following 'path' or '///' on the line are ok. Simple tests with the 'cql -parse' option suggest filter sequences such as 'path ... path ... /// ... path ... \n\n' or 'path ... /// ... \n ... path ... /// ... \n\n' or '/// "comment" \n' are allowed. '/// ... /// ... \n' is not allowed. Prior to CQL 6.2 '\n' was whitespace except for terminating a single line comment: now it is a terminator of filters where 'path' and '///' are involved.
RogerMarsh commentedon Feb 5, 2025
Commit d33c52f parses the examples supplied with CQL 6.0.4, CQL 6.1 and CQL 6.2, but does not handle the '--' (utf8 '―') filter, introduced at CQL 6.2, correctly.
The initial question is why are six '--' filters found in the babson1 example when only two are in the text.
The second question is why adjusting the babson1 example so the '--' filters are not in the body of a function causes the correct number of '--' filters to be seen (two not six).