Losing formatting other than what you’ve set in your deparser would be the point. Losing comments would be bad, but that seems easily fixable just by giving each comment block a symbol that points to it’s contents.
The best example I have is a closed source one and I can’t be more specific on what it is than to say that it’s probably installed on at least one of your Apple devices (assuming you have any).
Implementation-wise, the syntax tree nodes have additional attributes that hold pre- and/or post-element text. What’s on disk is the serialized tree. You edit a text version, and it’s parsed on every edit so it doesn’t have to be parsed again at evaluation time, and what’s stored is the parse tree with enough whitespace and comment hints to reconstruct the text for editing.
This is a case where looking at the textual code is rare, but hundreds of results must get updated in realtime on every change. This might be enough of a hint as to what program it is.
Wow, so this is possible.
Formatting is so damn arbitrary. Somebody has to have tried storing just the parse tree on disk, right?
The closest thing I’ve seen is Combobulate
If you do that, you lose formatting and comments every time you load the source from disk
Losing formatting other than what you’ve set in your deparser would be the point. Losing comments would be bad, but that seems easily fixable just by giving each comment block a symbol that points to it’s contents.
Or by including comments in the parse tree. (& Yes, it is done various places for various languages and formats.)
Do you have some examples?
(That is what I meant by giving them a symbol, maybe I worded it poorly)
The best example I have is a closed source one and I can’t be more specific on what it is than to say that it’s probably installed on at least one of your Apple devices (assuming you have any).
Implementation-wise, the syntax tree nodes have additional attributes that hold pre- and/or post-element text. What’s on disk is the serialized tree. You edit a text version, and it’s parsed on every edit so it doesn’t have to be parsed again at evaluation time, and what’s stored is the parse tree with enough whitespace and comment hints to reconstruct the text for editing.
This is a case where looking at the textual code is rare, but hundreds of results must get updated in realtime on every change. This might be enough of a hint as to what program it is.