We're currently having to anticipate some errors at a pre-parse stage--for example, the existing parser will happily replace properties it finds when a duplicate key, even if we think that should be an error (say when the text is missing an :end tag and it swallows the opening tag for a new post). We should fork the NYT parser to add additional strictness and throw errors when those problems occur.
Explicit markup for multiline text fields: field:: to mark their opening, and ::field to close them unambiguously. Inside this kind of block, no other fields are recognized.
Hooks for modifying keys and values on creation (for lower-casing, filtering)
Value parsing via a cast modifier (perhaps field<type>:)
We're currently having to anticipate some errors at a pre-parse stage--for example, the existing parser will happily replace properties it finds when a duplicate key, even if we think that should be an error (say when the text is missing an
:end
tag and it swallows the opening tag for a new post). We should fork the NYT parser to add additional strictness and throw errors when those problems occur.