For my purposes the +/ an +/ . * are superior functions with mapped
files. I do not understand how I could use such decoding with these in place?
I is fine that there is a way to improve J if there is commercial
demand for this feature in the future.
Anssi
At 18:05 4.10.2007, you wrote:
Depending on how the integers are used you may
have alternatives.
If you are using integers as ordinals or codes, then use
a literal matrix with a sufficient number of columns. e.g.
- encoding the age in years of people: using a 1-column
literal matrix (or a literal vector).
- dates as day numbers from 1960-01-01: use a 2-column
literal matrix (sufficient for about 180 years)
etc.
If not ordinals or codes, use a k-column literal matrix
where k e. 1 2 4 to encode k-byte integers, and convert
each time you need the numbers.
You have to decide is whether the extra programming
complications (or paying Jsoftware to implement 32-bit
integers in J64) are worth the savings in RAM.
----- Original Message -----
From: Roger Hui <[EMAIL PROTECTED]>
Date: Thursday, October 4, 2007 1:13
Subject: Re: [Jgeneral] Are 32bit integers in 64bit J possible?
To: General forum <[email protected]>
> It is theoretically possible for J64 to have k-bit integers
> where k e. 8 16 32 64.
> ----- Original Message -----
> From: Anssi Seppälä <[EMAIL PROTECTED]>
> Date: Thursday, October 4, 2007 0:44
> Subject: [Jgeneral] Are 32bit integers in 64bit J possible?
> To: General forum <[email protected]>
>
> Is it (even in theory) possible that 64 bit J could have 32 bit
> integers? The motivation is that integer mapped files of J64 require
> double size compared to J32. 1,5 GB J32 mapped is equivalent to 3 GB
> J64 mapped file and the GBs of RAM have a price.
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
