GF character encoding changes Thomas Hallgren 2013-12-18 ==Changes to character encodings in GF grammar files == Between the release of GF 3.5 and the next version, two changes were made relating to character encodings in GF grammar files: + The default character encoding was changed from Latin-1 (also known as iso-8859-1, cp1252) to UTF-8. + They way you specify alternate character encodings was changed. Instead of using a ``flags coding = ...`` declaration in the source file, you should now use a pragma ``--# -coding=...`` at the top of the file instead. == Advantages == UTF-8 is the default encoding for text files on many systems these days, so it makes sense to use it as the default for GF grammar files too. Changing how alternate encodings are specified allows conversion to Unicode to be done before parsing, which means that - we can allow Unicode characters in identifiers, not just in string literals, - it makes accurate column positions in error messages possible, - and (an implementation detail) we can use Alex to generate the lexer again. == How are my grammar files affected? == If your files still compile without errors after the change, you don't need to do anything. (But see Known problems below!) If you get one of the following errors, - ``lexical error``, - ``encoding mismatch``, - ``Warning: default encoding has changed from Latin-1 to UTF-8`` you need to add a ``--# -coding=...`` pragma to your file (or convert it to UTF-8). - For files containing only ASCII characters, no change is needed. - For files encoded in UTF-8 (and thus using a ``flags coding=utf8`` declaration), no change is needed. - For files containing Latin-1 characters (e.g. characters like å ä ö ü é), add a ``#-- -coding=latin1`` pragma at the top of the file. - For files using other encodings, copy the encoding specified in the ``flags coding=``//enc// to a corresponding ``--# -coding=``//enc//. Grammars will still compile with GF-3.5 after these changes. Note that GF only understands one option per pragma line. If you already have a ``--path=...`` pragma, you can not put the ``-coding=...`` option on the same line. Add it on a separate line: ``` --# -path=... --# -coding=... ``` The recommendation for the future is to use UTF-8 for all source files. == Known problems == The intention is that if a grammar file is affected by the changed default encoding, then you will see one of the messages listed in the previous section when you compile the grammar. But there are a couple if issues to be aware of: - Alex 3.0 seems to be confused about the length of matched strings sometimes. This can cause it to skip more than one line when it encounters a one-line comment in a grammar file with character encoding problems. So instead of a lexical error in the comment, you can get an odd syntax error on a subsequent line. - If you explicitly specify -coding=utf8 for a file that is not in UTF-8, you will not get an error, because the UTF-8 decoding function we currently use is forgiving, substituting the Unicode replacement character �, instead of reporting an error. Hopefully, we will be able to change this.