In Forth stacks are ubiquitous. Not only the data stack and the returnstack are used but many more can be found. Some of them hold configuration data like the search order stack which contains the wordlist-id’s for the interpreter. Amforth got the recognizers as an additional core level stack. All these stacks are placed in the EEPROM storage and they have a few things in common:
- they are used at system level.
- they are seldom changed.
- they are used with an iterator.
A stack is a contiguus eeprom space. The first cell has the actual stack depth, followed by the stack elements.
\ create a 11 elements stack > edp constant a-stack 12 cells eallot
a-stack is used to further work with the
There are three basic stack commands:
- get-stack and set-stack transfer the whole stack content to/from the data stack. That makes it possible to change the stack with the standard stack commands. Note, that the top-of-stack contains the actual stack depth.
- map-stack The iterator calls a predefined word for every stack element and leaves the iteration if the action word tells to do so
These commands transfer the data to/from the eeprom storage from/to the data stack. Only the actual stack depth is transferred.
The map-stack command is the stack iterator. It calls
an execution token for every stack element. The execution token
is expeced to return a flag to decide whether the iteration continues
or shall be ended prematurly. The command
map-stack itself leaves
a flag that informs about this termination cause. True means that
a premature exit has been done, false means that the iteration
was made for all elements.
The execution token is a nice example for Quotations. What it does is to use the stack element and generate a flag. If the flag is false (0), the data stack should be unchanged to make another iteration possible. If the flag is true (-1), the data stack can be changed to the final result.
The word called gets the actual stack element as the parameter. A flag is the return value. A true means, that this call was the last one, a false means, that the stack iteration continues with the next element.
A simple example is printing the word names of the recognizer stack. The EE_RECOGNIZERLISTLEN is a constant with the EEPROM address of the recognizer stack. The quotation extracts the name and prints it. The false flag makes sure, that every stack member is called. Since the final result of the iteration is not relevant, it gets simply dropped.
: .recs [: ( XT -- false ) >name icount $ff and itype space 0 ;] EE_RECOGNIZERLISTLEN map-stack drop ;
A slightly more complex iterator is the dictionary lookup word. It has to use the addr/len information for each wordlist from the ORDER stack. That makes it necessary to keep this information inside the quotation.
: find-name ( addr len -- xt +/-1 | 0) [: ( addr len wid -- xt +/-1 -1 | addr len 0 ) >r 2dup r> search-wordlist dup 0<> if >r nip nip r> -1 then ;] EE_ORDERLISTLEN map-stack 0= if 2drop 0 then ;
Since the quotation already deletes the addr/len from the data stack if the word is found, this cleanup is only necessary if no word could be found at all.
A similiar example is used for the recognizer stack. The main difference is the other meaning of the stack element and another iteration abort condition.
: do-recognizer ( addr len -- i*x r:table|r:fail ) [: ( addr len rec:XT -- i*x r:table -1 | addr len 0 ) rot rot 2dup 2>r rot execute 2r> rot dup r:fail = if drop 0 else nip nip -1 then ;] EE_RECOGNIZERLISTLEN map-stack ( -- i*x addr len r:table f ) 0= if \ no recognizer did the job, cleanup and add r:fail as default result 2drop r:fail then ;