Hi Blake, Elias,
OK, I changed this in SVN 278.
/// Jürgen
On 05/20/2014 07:29 AM, Blake McBride wrote:
I agree. I was surprised a couple of times when GNU APL automatically
terminated a string. I don't remember APL automatically terminating
strings. I think I'd rather see an error too.
Just an opinion.
Thanks.
Blake
On Tue, May 20, 2014 at 12:15 AM, Elias Mårtenson <[email protected]
<mailto:[email protected]>> wrote:
Hello Jürgen,
I know that you consider parsing unterminated strings to be a
feature, but I would ask you to reconsider this.
It is my personal opinion that this causes more opportunities for
confusion than potential benefits. I would like to share the
latest such problem that I came across:
I was looking at how GNU APL was parsing complex numbers, and I
typed the following:
* ⍎'2J3"
*
2J3
Note how I had accidentally terminated the string using a double
quote instead of a single quote. This was a typo on my part.
If unterminated strings had resulted in an error, as I am
proposing, I woul dhave gotten a SYNTAX ERROR (probably) and I my
mistake would have been clear. Instead, it looked almost correct.
I did notice that there was a space preceding the complex number,
so I did this:
* 8⎕CR ⍎'2J3"*
┌→──────┐
│2J3 ┌⊖┐│
│ │ ││
│ └─┘│
└∊──────┘
OK, now I was really confused. It took a while for me to figure
out that I had actually been bitten by the unterminated array
feature twice in a single statement: First, the input was parsed
by GNU APL as *⍎'2J3"'*. Then, the lamp function interpreted its
argument 2J3" as 2J3"", yielding a two-element array consisting of
a complex number and an empty array.
I think not giving an error in this case causes more confusion
than it's worth.
Finally, when reading the evaluation sequence in section 6.1.1 of
the standard, I interpret that as this is required to signal
syntax-error in this situation.
Regards,
Elias
