Re: cvthc not converting all values

[CRPence <CRPbottle@xxxxxxxxx>]

On 2/25/11 5:36 AM, Rick.Chevalier@xxxxxxxxxxxxxxx wrote:
> Thanks Barbara.  I hate to admit it but I hadn't considered
> negative numbers yet.  I'll take a look at LBCPYNV.
> The decimal positions shouldn't be an issue.
>
> On 2/24/11 4:01 PM, Barbara Morris wrote:
> > Rick doesn't have a packed variable that he can use with %EDITC.
> > He has a string with a packed value in it, where the length (and
> > decimals) are not known at compile time.
> >
> > Rick, if you know that you will never get a value with more than 31
> > digits, you could define a packed(63,31) variable in your program,
> > and use LBCPYNV to assign the packed value in your string to that
> > packed variable. Then you could use %CHAR on the packed variable.
> > That would handle negatives correctly. You'd probably want to trim
> > off trailing zeros if the input value has decimals.
> >
> > http://archive.midrange.com/rpg400-l/200605/msg00738.html

  Having started with the MI CVTHC, and the above archive message link 
showing the MI CPYNV [LBCPYNV], seems MI instructions are acceptable. 
Perhaps also consider using the MI EDIT [LBEDIT or EDITPD] instruction 
as an effective %EDITC or %EDITW.  Admittedly that requires using an 
edit mask, which could mean using a stored mask or having to generate 
the mask.

  Even if generating a mask is required, performing numeric editing 
directly could be better than moving the data into a fixed-size BCD 
[Binary Coded Decimal] scalar variable using CPYNV.  With EDIT the 
support is there for the maximum digits, precision and scale; i.e. not 
limited to the fixed-size(s) chosen.  If using the decimal scalar, 
further editing might be required to produce the desired effect, beyond 
the simple casting of a fixed-size decimal scalar to character; e.g. 
truncating trailing zeroes and possibly the decimal separator.  Also the 
original message had some "Format" and "Separator" information about the 
data for which an Edit Code or Edit Word might be appropriate; though 
those appear possibly to deal with date\time data versus numeric.

  The scenario was clarified in the message thread as:

> > > On 2/24/11 12:30 PM, Rick.Chevalier@xxxxxxxxxxxxxxx wrote:
> > > > Sorry, I wasn't clear. It is a character representation of a
> > > > packed field. The source value is a character field that looks
> > > > like x'12345F'. The length could vary as well as the decimal
> > > > places. I am 'unpacking' it into a character field that will
> > > > have '12345'.

  The EDIT can be used against numeric BCD data inline to character 
data by address, rather than requiring a [Packed decimal] scalar 
variable, just as with the use of CPYNV in the archived message; 
similarly by setting the attributes of the data at the given address. 
The results of my quick web search yielded an example of LBEDIT, but not 
an example of EDITPD; I verified a successful invocation of the former, 
but [could compile but] could not generate a valid instruction stream 
for the latter presumably due to a poor prototype.

  FWiW I believe the MI SETDP instruction optionally can be used to 
assign the data type, length and decimal places [precision and scale], 
and the address to the data, if a data pointer were to be used.  There 
seems no need for a DtaPtr, for use of the bound-program-access EDIT 
procedures _LBEDIT and _EDITPD.?

Regards, Chuck