error reporting - longest parse - count the lines - print the line - print the deepest non-alias node name - print the documentation of the node name - make it work with custom tokens read, with error reporting what was the bug with the large json from eskip? [next] missing tests, coverage: - validation - error cases - whitespace cases error reporting coverage custom tokens indentation streaming code generation go: - find things that depend on the syntax input - char matches can be generated into switches code generation js documentation flag support custom tokenization streaming verify choice and sequence preference formatter pretty report unused parsers go through the tests for docs test all docs warn nows usage in docs, e.g. spaces in symbol = [a-z]+ test error report on invalid flag report unused definitions simplify generated output parse hashed, storing only the results [optimization] try preallocate larger store chunks notes in formatting format: test back and forth equivalence of char classes format: test comments, first apply comments in the syntax [problems] can the root be an alias? check the commit mechanism [documentation] how the char classes are different from regexp why need nows when using ws