See also ClassifyForthCode
No Forth language standard provides the sort of regimentation that is expected of, say, Java. The latest Forth standard, approved by ANSI and ISO, allows a wide variety of Forth systems to be labeled standard and Forth code must be very carefully written if it is intended to be universally portable.
It would be possible to classify the various types of Forth implementations that are mentioned in the ANS standard, and create a sort of taxonomy of Forth systems. Code could be analysed by its needs and its portability might be somewhat defined.
Here is a third attempt, incomplete:
0 Memory00 address-unit size 000 other 001 8-bit 002 16-bit 01 char size 010 other 011 8-bit 012 16-bit 02 cell size 020 other 021 16-bit 022 32-bit 023 64-bit 03 endian ( mostly matters for C@ from cell data ) 030 other 031 little-endian 032 big-endian, cell same size as endian unit 033 big-endian, cell twice the size as endian unit 034 big-endian, cell half the size as endian unit 04 cell alignment 040 other -- can't ignore alignment but ANS words won't fix it 041 not aligned 042 aligned 05 float alignment 050 no floating point 051 not float aligned 052 float-aligned, double-float-aligned 06 harvard architecture ( code accessible? interferes with ALLOT etc?) 060 headers and/or code mixed with data 061 headers and code separate from data 07 Counted string representation 070 other 071 no counted strings 072 count is one character, followed by string 073 count is one character, followed by string, followed by null 074 count is another size, followed by string 075 count is another size, followed by string, followed by null 08 headers 080 other 081 no headers 082 header not accessible 083 '83 standard headers 09 I/O buffers 090 maximum size of keyboard input buffer 0900 other 0901 no buffer 0902 <80 chars 0903 <128 chars 0904 <1024 chars 0905 <32767 chars 091 maximum size of file input buffer 0910 other 0911 no buffer 0912 <128 chars 0913 <1024 chars 0914 < 32767 chars 092 maximum size of <# buffer 0920 other 0921 no buffer 0922 <2*cellsize+2 chars 0923 <128 chars 0924 <1024 chars 0925 < 32767 chars 0A Other buffers 0A0 PAD 0A00 other 0A01 no buffer 0A02 <84 chars 0A03 <1024 chars guaranteed 0A04 all unused memory 0A1 blocks 0A10 other 0A11 no buffer 0A12 1 block buffer
1 Arithmetic10 arithmetic type 100 other 101 2's-complement 102 1's-complement 103 sign-magnitude 11 division type 110 other 111 no division 112 symmetric 113 floored
2 Return stack ( system stuff interferes with use of stack? Change system data?)20 access to "real" return stack? 200 other 201 no access to real return stack 202 nest-sys available with R>, size of cell 203 nest-sys available, double size of cell 204 nest-sys available, half size of cell 21 where is loop-sys? 210 other 211 no loop-sys 212 loop-sys on return stack, uses 2 cells 213 loop-sys on return stack, uses 3 cells 22 switch from interpret to compile implemented as 220 other 221 deferred word 222 branch 223 call
3 Data stack ( which system behaviors interfere with user data stack?)30 where is control-flow stack? 300 other 301 no control-flow stack 302 control-flow stack on data stack 31 where is colon-sys 310 other 311 no colon-sys 312 colon-sys on data stack
4 The state-smart mess40 How many xt's can a word have? 400 other 401 one 402 two, one for interpreting and one for compiling
5 Particular words50 Do BEGIN loops allow more than one WHILE? 500 no BEGIN 501 no 502 yes 51 What does DOES> apply to? 510 other 511 no DOES> 512 the last definition made by CREATE 513 the last named definition 514 the last definition 52 What does IMMEDIATE apply to? 520 other 521 no IMMEDIATE 522 the last named definition 523 the last definition
6 True flag60 What value do primitives return for a true flag? 600 other 601 no true flag 602 low bit set 603 all bits set 61 What value do primitives accept for a true flag? 610 other 611 no true flag 612 low bit set 613 any bit set
7 Is floating point available?70 no floating point 71 yes, with floats on data stack 72 yes, with separate floating point stack of less than 17 items 73 yes, with separate large floating point stack
8 Permanent data storage80 other 81 none 82 direct Access by Blocks 83 file System
9 Graphic character types
90 other 91 no characters 92 ascii7 93 ISO8 94 standard 16-bitA Number representationA0 maximum base A00 no base A01 other A02 maximum BASE = 36 A03 maximum BASE = 62 A1 <# location ( for nested number output, mostly done when debugging) A10 no <# A11 other A12 a single fixed buffer A13 a single transient buffer A14 chars stored on data stack
I'm supposing I'd never need to know about names longer than 31 chars
there may be an additional count issue, related to string length: does count fetch 8 bit, or 1 chars ? maybe the question should be "how wide is 1 CHARS", assuming that COUNT fetches string length as well as chars from a string.
Yes, that's the issue. If the count is the same as a character and a counted string has the count right before the string, then you can use COUNT to step through the string. If either of those isn't true, you can't.
There's also a question how big a counted string can be.
according to DPANS, number conversion must be able to work with base in the range of 2...36. it is an implementation issue whether base >36 still gives meaningful results.
Good point!
