CoCoA » CoCoALib

The input functions are quite messy: not really a coherent design philosophy.

It seems that ReadExpr gives error if there are extra chars after the end of the expr, but ReadExprSemicolon does not complain if there are extra chars after the semicolon.
At the moment my expectation is that a fn which accepts a stream would read the next item on the stream, and leave the input pointer at the first char after the read value.
If the caller wants to ensure that there is nothing else in the stream then this must be written explicitly in the caller's code.

With this design ReadExprSemicolon becomes even simpler to implement.

Maybe there could be ReadJust1Expr which behaves as ReadExpr currently does?

I have just checked what the code does (or seems like it should do).
The current impl is:

• reading from istream keeps going until an "impossible" char is found; error if expr up to that point is incomplete, o/w no error & remaining chars are ready for further input operations
• reading from string expects the string to contain exactly the right input: trailing non-whitespace chars will trigger an error

I have changed the CoCoA-5 parser so that it produces a warning when given input such as 2*-1 or 3/-1. The computation still works, but the warning recomments putting the 2nd arg inside brackets. Also 2*+1 and 3/+1 produce the same warning.

An impromptu test suggests that all works as desired.