Seperate code from data

[mitra42]

There are a number of places where code and data are mixed, preventing them being separated - e.g. in Rom.

This is a feature of many other Forth's.

Specific known issues ...

• [x] VARIABLE
• [x] Vocabulary words - plus the hand-constructed "FORTH"
• [x] Anything built with CREATE
• [ ] What else?

[mitra42]

I'm working on this in the eprom branch

[mitra42]

My current assumptions are:

• a memClass will separate addresses into writable and not-writable areas of memory.
• writes to Code or Name dictionary are OK since will happen before building EPROM and that then there will be live Code/Name dictionaries in writable space.

It follows from those assumptions:

• [x] we can build a memClass that enforces these assumptions during testing e.g. using top bit of address.
• [x] the only places that matter are non-defining writes to Code or Name space.
• [x] possibly mistaken assumptions of where PAD or HLD are relative to CP (TIB is below Return so OK)
• [x] possibly places writing with encodeString

Maybe No changes to

• [x] CREATE defines I/O and NULL$ and - 'BOOT which aren't ever written, BUT make sure uses correct CREATE
• [x] ! C! +! 2! cStore() store()
• [x] PACK$ it copies to NP but NP will be moved to space it can write into
• [x] Other string ops CMOVE PACK$ FILL WORD (uses PACK$) $," : no obvious misuse
• [x] TIB - is below Return Stack
• [x] Mstore, Mstore8, Mpush, Mpush8 (any obvious misuse is flagged)
• [x] m.encodeString `- [x] JStoCounted - see below
• CODE - only used when defining
• JStoTIB goes into TIB correctly

Next steps rework (callers are all labelled)

• [x] , -> v, (tricky as writes back)
• [x] CREATE - note defined two places -> vCREATE
• [x] VARIABLE
• [x] watch for tokenVar (as used by VARIABLE) that exposes a writable address inside a dictionary
• [x] ALLOT -> vALLOT - tricky as stores back
• [x] cpFetch -> vpFetch
• [x] CPoffset -> VPOffset
• [x] CP -> VP
• [x] Add VP to User vars (see use of CPoffset)
• [x] padPtr use vpFetch
• [x] PAD to use vHERE
• [x] HERE -> vHERE
• [x] jsToCounted use vpFetch
• [ ] .FREE -> .vFREE and check usages

Rework these that update vocab

• [x] FORTH definition at start of dict
• [.] OVERT (code and Forth)

Check for usages of:

[mitra42]

This is mostly done - in the "eprom" branch.

The open questions are how to handle defining after any Eprom interaction - and that might need to be an exercise for the future. Specifically, if we can define then we need to move "FORTH" into the data space (which is slightly non trivial) and define tokenVocabulary to use data space, and make sure to initialize CP and NP to point to wherever you can define when half the dictionary is in ROM.

[mitra42]

This is done - both for Arduino and in Javascript either in a single memory space (MemNN_MM) or separate (Rommable16_16)