Question
Eduard Lebedyuk · Apr 15, 2019

Iterate over %List/$lb in C

I'm using callin to get global values.

Here's a simple function to get string value from global and return it:

int GetGlobalStr(char *global, CACHE_EXSTRP result)
{
    int push = CACHEPUSHGLOBAL(strlen(global), global);

    // narg Number of subscript expressions pushed onto the argument stack.
    int narg = 0;

    // flag - Indicates behavior when global reference is undefined:
    // 0 — returns CACHE_ERUNDEF
    // 1 — returns CACHE_SUCCESS but the return value is an empty string.
    int flag = 1;

    int get = CACHEGLOBALGET(narg, flag);

    int pop = CACHEPOPEXSTR(result);

    return ZF_SUCCESS;
}

I get global value in result successfully. However I need to iterate over $lb. How can I do that?

#Beginner #C++ #Callout #Caché
00
2 0 11 305
Log in or sign up to continue

Replies

Robert Cemper · Apr 15, 2019

the structure of $LB() is rather simply a binary string

-----------element--------------
TotalLength = 1, 3, 7 bytes depending on size    *corrected*
Type = 1 byte  (check in JSON converter for codes, or just check with ZZDUMP)
Content : size = TotalLength-1-size of length field
-----------element--------------
TotalLength = 1, 3, 7 bytes depending on size    *corrected*
Type = 1 byte  (check in JSON converter for codes, or just check with ZZDUMP)
Content : size= TotalLength-1-size of length field
-----------element--------------

...

Therefore concatenation of $lb) is so easy

30
Robert Cemper · Apr 15, 2019

special case:

s a=$lb() zzdump a w !,$l(a)
 0000: 01                                                      .
1
s a=$lb("") zzdump a w !,$l(a)
 0000: 02 01                                                   ..
2
s a="" zzdump a w !,$l(a)

0
20
Robert Cemper · Apr 15, 2019

the total size is somewhat strange its format changes from 255 to 256 in size and interpretation
and again at 65535 / 65536   up to <MAXSTRING>

 

10
Eduard Lebedyuk · Apr 15, 2019

Callin is stack-based. You push function arguments to a stack, call function, read result from stack.

These two functions are used to push and pop $lb to stack, not to get individual $lb elements.

00
Eduard Lebedyuk · Apr 30, 2019

Thank you, @Steven Hobbs ! I'll try to fix my code on LE platforms at the very least.

There are some unusual rules that must be followed if you want to get the same results as the $LISTxxx functions get in IRIS/Caché.

Is there any place I can familiarize myself with these rules?

00
Eduard Lebedyuk · Apr 15, 2019

Thank you Robert!

To be honest I'm hoping someone has C code to share.

00
Dmitry Maslennikov · Apr 15, 2019

So, you can convert $List to some kind of List in C?

$LIST is a very simple format, but you can't get any particular item just by position. You should go through the list from the first item.

In a simple explanation, it is just concatenation of individual $LIST. So, you can't say how many items you have until you go through this list and count them.

00
Vitaliy Serdtsev · Apr 15, 2019

See %CACHE_HOME%\dev\Cache\callout\demo\czf.pdf (Section "Lists")

00
Eduard Lebedyuk · Apr 18, 2019

Here's a sample C code to iterate over $lb structure.

#include <math.h>

/// Convert unsigned integer bytes to integer
int64_t makeint(const char *buff, size_t offset, size_t offsetinint, size_t len)
{
    union
    {
        int64_t i64;
        uint8_t u8[8];
    }d64;

    offsetinint = offsetinint & 7;
    memset(&d64, 0, sizeof(d64));

    memcpy(&d64.u8[offsetinint], buff + offset, len > (8 - offsetinint) ? (8 - offsetinint) : len);
    return d64.i64;
}

/// get next power of 2 greater than v
int64_t next2(int64_t v)
{
    v--;
    v |= v >> 1;
    v |= v >> 2;
    v |= v >> 4;
    v |= v >> 8;
    v |= v >> 16;
    v |= v >> 32;
    v++;
    return v;
}

/// List types. NONE and NONE2 are placeholders
/// See %CACHE_HOME%\dev\Cache\callout\demo\czf.pdf (Section "Lists")
enum ListTypes {NONE, STRING, USTRING, NONE2, INTP, INTN, DOUBLEP, DOUBLEN, FLOAT};

/// This function iterates over all elements in a list
void ListToTuple(CACHE_EXSTRP result)
{
    // $lb structure
    char* list = result->str.ch;
    int listLength = result->len;

    // current element
    int num = 0;

    // current byte position
    int i=0;

    // length of current element
    int l = 0;

    // datatype
    int type = 0;

    while (i<listLength) {

        // Calculate length of current element - START
        if (0 == (l = (list[i]&255))) {

            // First BYTE is 0, length is in following 2 BYTEs
            size_t t_n = ((list[i+1]&255)|((list[i+2]&255)<<8));
            if (t_n != 0) {
                  l = t_n + 3;
            } else {
                // 4 Byte length
                l = ((list[i+3]&255) | ((list[i+4]&255) << 8) | ((list[i+5]&255) << 16) | ((list[i+6]&255) << 24)) + 7;
            }
        }
        // Calculate length of current element - END

        // Calculate data position - START
        int dataStart = 0;
        int dataLength = 0;

        if (l < 255) {
            type = list[i+1];
            dataStart = i + 2;
            dataLength = l - 2;
        } else if (l < 65536) {
            type = list[i+3];
            dataStart = i + 4;
            dataLength = l - 4;
        } else {
            type = list[i+7];
            dataStart = i + 8;
            dataLength = l - 8;
        }
        // Calculate data position - END
        
        if (type==STRING) {
            char* value;
            memcpy(value, list+dataStart, dataLength);
        } else if (type==USTRING) {
            char* value;
            memcpy(value, list+dataStart, dataLength/2);
        } else if (type == INTP) {
            int64_t value = makeint(list, dataStart, 0, dataLength);
        } else if (type==INTN) {
            int64_t value = 0;
            if (l==2) {
                value = -1;
            } else {
                memcpy(&value, list+dataStart, dataLength);
                if (value == 0) {
                    value = - (1 << (dataLength * 8));
                } else {
                    int64_t pow2 = next2(value);
                    value = value - pow2;
                }
            }
        } else if (type==DOUBLEP) {
            int64_t temp = makeint(list, dataStart+1, 0, dataLength-1);
            signed char exp = list[dataStart];
            double value = temp*pow(10, exp);
        } else if (type==DOUBLEN) {
            int64_t temp = 0;
            memcpy(&temp, list+dataStart+1, dataLength-1);
            if (temp == 0) {
                temp = - (1 << (dataLength * 8));
            } else {
                int64_t pow2 = next2(temp);
                temp = temp - pow2;
            }
            signed char exp = list[dataStart];
            double value = temp*pow(10, exp);
        } else if (type==FLOAT) {
            double value;
            memcpy(&value, list+dataStart, 8);
        }
        
        i += l;
        num++;
    }
}
00
Accepted answer
Steven Hobbs · Apr 25, 2019

There are some issues with above C code for manipulating a string of $LIST elements.

I do not believe the above code will work on a big-endian platform, such as PowerPC running the AIX operating system.  Conversions from $LIST representation to numeric representation will order the bytes backwards on big-endian hardware.

Decimal floating-point values in IRIS and Caché have almost 19 decimal digits of precision while the above code translates these numbers to IEEE binary double-precision floating-point values which have less than 16 decimal digits of precision.  This means that $LISTBUILD(0.3) will be converted by the above code into the value 0.29999999999999998889... .  The above code also introduces a double-round error so that decoding $listbuild(x) and $listbuild($double(x)) will not always be equal because the $double(x) function in IRIS/Caché will do the conversion from decimal to binary without a double round.

The above code is inconsistent in resulting type of integer values.  Consider,
USER>set L5A=$LISTBUILD(5),L5B=$LISTBUILD(50/10)
USER>WRITE $LISTSAME(L5A,L5B)
1

The $LIST elements in L5A and L5B contain the same value, 5.  However, the above code will convert L5A to the C int64_t type while it will convert L5B to the C double type.  If the value of the integer is greater than 2**53 then the above C code can convert the identical integer values into different integer values.

Also, the above code does not correctly handle all the special cases when (type==FLOAT) is true.

The IRIS/Caché $LIST representation is not as simple as most people think.  There are some unusual rules that must be followed if you want to get the same results as the $LISTxxx functions get in IRIS/Caché.
 

10