On 19-Apr-2015 12:26 -0500, Hoteltravelfundotcom wrote:
I have coded this:
eval ord_wt = %dec(PHWtot:7:2)
but still got error of data decimal error.
To locate this I have taken a D' and have the program dump.
Is this all I need? here it will tell me which is the bad data?
From the dump data showing the value of variable PHWtot [when RNX0105
was the result], one should be able to infer what "is the bad data".
What a valid character string representing a _decimal value_ is [at
least somewhat] documented in the RPG reference:
<
http://www.ibm.com/support/knowledgecenter/api/content/SSPSQF_9.0.0/com.ibm.etools.iseries.langref.doc/evfrilsh828.htm#bbdecnum>
Another definition [that includes overflow RNX0103 as /invalid/ too],
for what %DEC can handle in converting a character [string] to a numeric
value, might be [and IMO is more accurately described by]:
• zero to many leading blanks
• an optional leading sign character as either the plus ('+')
character for a positive value or the minus ('-') character for a
negative value
• zero to many blanks
• zero to many decimal digit characters zero ('0') [surely at least
as many as the maximum number of digits supported precision of 31 or 63,
though possibly not restricted except by size of string]
• zero to many decimal digit characters from zero ('0') to nine ('9')
[to avoid overflow, only for as many as, up to the result of the
expression (third_argument - second_argument) of the %dec built-in; i.e.
the expression (precision-decimal_places), which in the given scenario,
is the expression (7-2) derived from the %dec(arg1:7:2), for a result
of: /as-many-as 5 digits/]
• an optional decimal separator [aka decimal point] as either the
period ('.') character or the comma (',') character
• zero to many decimal digit characters from zero ('0') to nine
('9'), unlimited in number except by the length of the character string
• zero to many blanks; i.e. as insignificant fractional precisions,
they are meaningless to the outcome and can not contribute to an
overflow condition like the integer portion can
• an optional trailing sign character as either the plus ('+')
character for a positive value or the minus ('-') character for a
negative value, but only if there was _no_ leading sign
• zero to many trailing blanks
Otherwise can I code in the RPG to handle such an occurrence?
Uh, sure. The RPG could be coded to do anything to /handle/ the
signaled\monitored problems that can occur. But depending on what the
definition of /handle/ is in that sentence, the problem could be
intractable; how an occurrence of bad data might be /handled/, is
necessarily _specific_ to both the application and the requirements to
maintain the integrity of the [calculated and stored] results. For
example, a conspicuously poor /handling/ for a business application,
would be setting a recorded price to zero whenever the price being input
as character, was deemed invalid according to the %dec casting.
midrange.com
