The premise is false. Copies of the inputs are rarely made, usually
only when conversion is required. When possible, the argument is
operated on inplace with the result occupying the same memory as the
argument.
Art does have a point about the Nub documentation, but it is important
that the opening lines of the NuVoc entry focus on answering the
question "What does ~. y do?" without repeating boilerplate to the
effect that the result is distinct from the argument. Nobody expects -5
to change the value of 5, or ~.a to change the value of a. IMO that
entry is OK as it stands.
Henry Rich
On 6/9/2022 6:09 AM, Arthur Anger wrote:
All--
Each non-assigment step in a computation produces copies(!) of its inputs,
selected and arranged in various specified ways, awaiting disposition--in
another step, in assignment to storage, in output, or in discard.
At the very least, any recipe for processing an array should begin with some version of
"Take a copy of y, then...", and consistently refer to that copy, rather than
to y.
As an example, the page for Nub currently contains the following statements:
Removes duplicates from a list
<https://code.jsoftware.com/wiki/Vocabulary/AET#List>.
More generally, removes any item
<https://code.jsoftware.com/wiki/Vocabulary/AET#Item> of an array
<https://code.jsoftware.com/wiki/Vocabulary/AET#Array> that matches a preceding item.
1. Remove duplicate values
<https://code.jsoftware.com/wiki/Vocabulary/Glossary#Value> from a lis
<https://code.jsoftware.com/wiki/Vocabulary/AET#List>t
2. Remove duplicate rows from a table
<https://code.jsoftware.com/wiki/Vocabulary/AET#Table>
I recommend replacing them with:
Copies items from a list, omitting duplicates.
More generally, omits any item
<https://code.jsoftware.com/wiki/Vocabulary/AET#Item> of an array
<https://code.jsoftware.com/wiki/Vocabulary/AET#Array> that matches a preceding item.
1. Omit duplicate values
<https://code.jsoftware.com/wiki/Vocabulary/Glossary#Value> from a lis
<https://code.jsoftware.com/wiki/Vocabulary/AET#List>t
2. Omit duplicate rows from a table
<https://code.jsoftware.com/wiki/Vocabulary/AET#Table>
The language used to describe a verb often suggests that its result will
somehow replace the values being operated on, which will be (nearly) accurate
only if followed immediately by re-assignment to the same, named, argument.
I urge that any new documentation observe this distinction, and that any
definition being updated follow this pattern. If there is a consensus that
this is a worthy cause, I shall be willing to propose such changes where I may
encounter them in coming months.
--Art
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
--
This email has been checked for viruses by AVG.
https://www.avg.com
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
