problems inserting records into an array

[Travis Siegel]

I'm porting a program from windows to dos (yes, I know, odd, but true) I've got most of it compiling, after writing a couple functions to replace missing dos ones, to keep source code changes to a minimum.
However, my current problem is that I can't get array insert to work properly.
I'm getting an error 420, Relational operator expected.
I have an array of records, and I'm trying to insert a record into the array.
The line of code in question currently reads as:

array insert varlist(varindex),temp3

Temp3 is a record of a user defined type.

type var1
vname as string*40
vvalue as string*40
end type

The original lines of code from the windows code was:

varlist(varindex).vname = temp1
varlist(varindex).vvalue = temp2

On windows, the values are inserted directly into the array automatically.
On dos, apparently, I need to setup the record first, then insert it as a complete set.
I'm getting errors which seem to indicate I need to define the type of item I'm inserting.
Normally, I'd just add a $ % #, or whatever is necessary to identify the datatype being inserted into the array. However, in this case, since it's a record, there isn't really any type identifier (that I know of anyway), so what can I do here to get this record inserted into the array?
Any suggestions/assistance would be much appreciated.

[Michael Mattias]

Original code is not doing an ARRAY INSERT as far as I can tell, it's simply doing a replace of the element. Of course, what's not shown may make a liar of me.

But bottom line, ARRAY INSERT works the same on PB/DOS as it does on PB/Windows. In either case, the replacement element (temp3) will be inserted "as it currently exists" meaning if it ain't been initialized with the correct values, it will go in in its default condition, which depends on its scope and prior use in the program and/or procedure. (Code not shown).

MCM

[Travis Siegel]

[QUOTE=Michael Mattias;305371]Original code is not doing an ARRAY INSERT as far as I can tell, it's simply doing a replace of the element. Of course, what's not shown may make a liar of me.

But bottom line, ARRAY INSERT works the same on PB/DOS as it does on PB/Windows. In either case, the replacement element (temp3) will be inserted "as it currently exists" meaning if it ain't been initialized with the correct values, it will go in in its default condition, which depends on its scope and prior use in the program and/or procedure. (Code not shown).

But, this is my problem, the array insert is not working, I get an error (mentioned above) when trying to do the insert.
In the windows program, the array is initially dimensioned to 10 records, then another 10 records are added each time it runs out of usable records. Are you saying the same behavior is possible under dos, and therefore no array insert is necessary?
I was unable to get this to behave, though I can certainly try again using the same approach used under windows, (I was under the impression you *had* to do an array insert to get the record in, but if modifying existing blank entries works, then I'll just do that, it will make things easier.
But, I'm still puzzled why the array insert gives an error, but I guess it's not important if the other approach works.

[Michael Mattias]

the array is initially dimensioned to 10 records then another 10 records are added each time it runs out of usable records. Are you saying the same behavior is possible under dos, and therefore no array insert is necessary?

Yes.

REDIM PRESERVE is used to change the bounds of an array without discarding the existing elements.

ARRAY INSERT/DELETE are used to "shift" elements within the current bounds, discarding the deleted element (ARRAY DELETE) or the element at the current upper bound (ARRAY INSERT).

PB-DOS or PB-Windows, works the same. So, do you want to insert elements within the currrent array bounds or resize the array to hold additional elements?

Under PB/DOS you have some restrictions when the total memory requirements for all array elements exceeds 128Kb, but other than that the code should be identical.

MCM

[Travis Siegel]

[QUOTE=Michael Mattias;305383]Yes.

REDIM PRESERVE is used to change the bounds of an array without discarding the existing elements.

ARRAY INSERT/DELETE are used to "shift" elements within the current bounds, discarding the deleted element (ARRAY DELETE) or the element at the current upper bound (ARRAY INSERT).

PB-DOS or PB-Windows, works the same. So, do you want to insert elements within the currrent array bounds or resize the array to hold additional elements?

Under PB/DOS you have some restrictions when the total memory requirements for all array elements exceeds 128Kb, but other than that the code should be identical.

I'm aware of all of this, but several things are an issue here.
1. code is not identical, the windows code for adding records to the array did not work, it gave an error.
2. adding the values to the record itself work fine, inserting that record into the array does not.
3. trying to modify the entries in the array directly doesn't work either.


Let me try again to explain what's happening (or not as the case may be)

The array is made up of records.
The records are made up of two string variables, both of which are 40 chars long.
As mentioned before, when trying to insert a record into the array, pbdos gives me an error (already reported above), and I also get an error when trying to assign values to the record in an empty array as well.
The windows code was something like this: (don't have it here, so can't copy/paste)

varlist(varindex).vname = temp1
varlist(varindex).vvalue = temp2

This code does not work in dos, it gives an error saying it's expecting a comma where the period is placed.
Changing the code to read

temp3.vname = temp1
temp3.vvalue = temp2
varlist(varindex) = temp3

sdoesn't work either.
In this case, temp3 has been identified as type var1, which is what varlist is an array of (I.E. varlist is an array of var1 types)
I'm obviously missing something, but I can't for the life of me figure out what.
Why is the array insert giving an error?
Why does the varlist(varindex).varname assignment give an error, what's the syntax for assigning a value under dos (it's not identical to the windows syntax as already mentioned above), so what is the proper syntax then?
The dos help seems to indicate that at least one of these approaches should work, and as far as I can tell, it should, but it's expecting a type identifier such as $ # % or something else to tell it what kind of element the temp3 is.
And, another (slightly) related question.
Why do I need to identify some variables as strings and sometimes not.

temp1 as local string, temp2 as local string
temp1 = "hello"
temp1 = temp2

sometimes works, and sometimes doesn't. Sometimes it's requiring the $ symbols on the temp1 and temp2 identifiers, but not always.
Anyone know why this is, and how to make it consistent?

[Michael Mattias]

If it's this statement here causing a compile-time error....

array insert varlist(varindex),temp3

...and temp3 is the same datatype as array varlist() (insufficient code shown), then you have a support issue and should contact PB support, because that should work to insert "temp3' as the new "varindex-th" element of "varlist()", shifting all element numbers greater than or equal to varindex "up by one", discarding the element at varlist (UBOUND(varlist)).

MCM

[Paul G. Penn]

Version of PB/DOS being used

I'm not sure about something. What version of PB/DOS are you using for the rollback? (I'm wanting to go in the other direction.) I had the same problem and asked PB Support. I'm using PB/DOS Version 3.0c and was told that the ARRAY statement does not support UDTs. That could be your whole problem. You might have to go to PB/CC. An extreme solution would be to convert the UDT to a flex string, add the element, then convert back. The ARRAY statement should work normally when the array is of a PB predetermined type (BYTE, WORD, INTEGER, etc.) I'm not sure about Version 3.5. I was told PB/CC (and PB/Win) had added a custom feature to the ARRAY statement that would support UDTs. Just something else to consider.

[Michael Mattias]

compiles fine with PB/DOS 3.5

TYPE FOO
   A AS STRING * 40 
   B AS STRING * 40 
END TYPE
FUNCTION  TESTER() AS INTEGER    
 REDIM F (10) AS FOO  
 DIM Foo1 AS FOO  
 ARRAY INSERT F(5), Foo1  
END FUNCTION

Can't tell if it's working. MY PB/DOS stuff is in some kind of security lockdown on my system and I can't unset the 'read only' attribute in the working folder.

But 'compile' gets as far as "desitnation file write error" which means the compiler liked it and TRIED to write the EXE...

MCM

[Arthur Gomide]

a simple test

TYPE FOO
   A AS STRING * 4
   B AS STRING * 4
END TYPE
FUNCTION  TESTER() AS INTEGER
   REDIM F (10) AS FOO
   DIM Foo1 AS FOO
   F(10).A = "xxxx"
   F(10).B = "XXXX"
   PRINT F(10)
   Foo1.A = "alfa"
   Foo1.B = "beta"
   ARRAY INSERT F(5), Foo1
END FUNCTION
DIM F(9) AS FOO
FOR I = 0 TO 9
   J = I+48
   F(I).A = STRING$(4,J)
   F(I).B = STRING$(4,CHR$(J+16))
NEXT I
CLS
FOR I = 0 TO 9
   PRINT I;">";F(I).A;F(I).B
NEXT I
PRINT : tester : PRINT
FOR I = 0 TO 10
   PRINT I;">";F(I).A;F(I).B
NEXT I
END

a wrong result

0 >[email protected]@@@
1 >1111AAAA
2 >2222BBBB
3 >3333CCCC
4 >4444DDDD
5 >5555EEEE
6 >6666FFFF
7 >7777GGGG
8 >8888HHHH
9 >9999IIII

xxxxXXXX

0 >[email protected]@@@
1 >1111AAAA
2 >2222BBBB
3 >3333CCCC
4 >4444DDDD
5 >5555EEEE
6 >6666FFFF
7 >7777GGGG
8 >8888HHHH
9 >9999IIII
10 >yz

]

[Michael Mattias]

Invalid test. F() is not shared with function TESTER, which uses a LOCAL array.

[Arthur Gomide]

Sure!

TYPE FOO
   A AS STRING * 4
   B AS STRING * 4
END TYPE
DIM F(9) AS FOO
FOR I = 0 TO 9
   J = I+48
   F(I).A = STRING$(4,J)
   F(I).B = STRING$(4,CHR$(J+16))
NEXT I
CLS
FOR I = 0 TO 9
   PRINT I;">";F(I).A;F(I).B
NEXT I
PRINT

REDIM PRESERVE F(10)
DIM Foo1 AS FOO
Foo1.A = "alfa"
Foo1.B = "beta"
ARRAY INSERT F(5), Foo1
PRINT

FOR I = 0 TO 10
   PRINT I;">";F(I).A;F(I).B
NEXT I
END

result

0 >[email protected]@@@
1 >1111AAAA
2 >2222BBBB
3 >3333CCCC
4 >4444DDDD
5 >5555EEEE
6 >6666FFFF
7 >7777GGGG
8 >8888HHHH
9 >9999IIII

0 >[email protected]@@@
1 >1111AAAA
2 >2222BBBB
3 >3333CCCC
4 >4444DDDD
5 >alfabet <<<< ????
6 >5555EEEE
7 >6666FFFF
8 >7777GGGG
9 >8888HHHH
10 >9999IIII

[Michael Mattias]

Same results here but...

that triggers a memory.... I seem to recall losing a character on ARRAY INSERT of UDT arrays like this was a known bug in some version(s?) of PB/DOS

Mine:

09/07/1998  11:44 AM           350,731 PB.EXE

Contact support.

[Arthur Gomide]

If so, my copy has that bug!

19/12/1997  03:50           350.827 PB.EXE

[Michael Mattias]

Fortunately, there is a VERY simple workaround:

TYPE var1
    vname as string*40
    vvalue as string*40
[B][COLOR="Red"]    unused as integer [/COLOR][/B]
end type

[Arthur Gomide]

Another solution for circumvent the bug:

TYPE FOO
   A AS STRING * 4
   B AS STRING * 4
END TYPE
...
ARRAY INSERT F(5) : REM , Foo1

DEF SEG = VARSEG(F(5))
POKE$ VARPTR(F(5)), Foo1
DEF SEG
...

or only

TYPE FOO
   A AS STRING * 4
   B AS STRING * 4
END TYPE
...
ARRAY INSERT F(5) : REM , Foo1
F(5) = Foo1
...