INMC 80 News

  

October–December 1981, Issue 5











Page 64 of 71











The use of manifest constants at the top of the program greatly helps the alteration of it in the future, the conversion of it to other systems, which may have different parameters to yours (think of the T2/T4/BBUG monitors use of different control characters), and considerably improves the legibility and hence the clarity of the program. It also, at the expense of a slightly longer assembly time, ensures that any changes are correctly installed in the program. If your assembler supports multiplication and division, it can also calculate displacements up and down tables, and computer constants, so that they can be manifested in this way. For example, in preparing a table to hold 4 byte entries, one can write:

NUMENT:EQU 131 ;Number of entries expected

ENTSIZ:EQU 4   ;length of individual entry

TABLE:DEFS ENTSIZ*NUMENT

and if the entry length or the number of entries is changed, the assembler makes the change automatically. Should the assembler not support multiplication or division, one could write

TABLE: DEFS NUMENT+NUMENT+NUMENT+NUMENT

This is a bit more cumbersome, but it does clearly show how the size of table is allocated, particularly if accompanied by a comment to the effect that each entry is four bytes long.

We haven’t spoken of BASIC for some time. Has this business of manifest constants anything for that? Yes! BASIC supports a considerable number of variables, which are a letter, followed by a letter digit or space, to give a larger number than could reasonably be exhausted in the course of a program. It is possible to define a number of variables at the start of a program such as:

100 CR=13:BS=8:FF=12:REM values for carriage return, b/s, and form feed

Similarly, if testing for a value in a list, one could define:

110 YES=1:NO=0

and use these as manifest constants. This considerably improves the legibility of the program, and anything which does that in BASIC has to be a good idea! It also improves the running speed, as BASIC stores all constants in ASCII internally, and each time it comes across one, it has to translate it into binary, but each variable is already translated ready to go! Notice also that it is possible, at the expense of a slight increase in running time, to use long variable names. Note however that BASIC only uses the first two characters of the name to identify it. This means that FIRST and FIZZ are the same to BASIC. An easy way out is to put a unique padder in front of the meaningful part of the name (meaningful to you, but not to BASIC) – say Z1FIRST and Z2FIZZ. It becomes very easy to disregard the leading characters, which make the variable distinguishable to BASIC, but you the human look mainly at the end.

In “Writing Interactive Compilers and Interpreters”, P.J. Brown says “Thus some compiler writers say that a compiler should contain no numbers at all, except perhaps, as a special treat, the occasional zero or one.” (p69). While this may be a little extreme, it is none the less a good guideline.


This is an OCR’d version of the scanned page and likely contains recognition errors.











Page 64 of 71