This chapter describes all of the FML and FML VIEWS functions, with the exception of the run-time mapping functions described in Chapter 4. In this chapter you will learn:
For COBOL programs, the FML functions are not directly
available. A procedure, FINIT is available to
initialize a record for receiving FML data, and FVSTOF
and FVFTOS are available to convert from a COBOL
record to an FML buffer and back. These are described in detail
in the BEA TUXEDO COBOL
Guide. The COBOL interface will not be described further
in this chapter.
There are two variants of FML. The original FML interface is based on 16-bit values for the length of fields and containing information identifying fields (hence FML16). FML16 is limited to 8191 unique fields, individual field lengths of up to 64K bytes, and a total fielded buffer size of 64K. The definitions, types, and function prototypes for this interface are in fml.h which must be included in an application program using the FML16 interface; and functions live in -lfml. A second interface, FML32, uses 32-bit values for the field lengths and identifiers. It allows for about 30 million fields, and field and buffer lengths of about 2 billion bytes. The definitions, types, and function prototypes for FML32 are in fml32.h; and functions live in -lfml32. All definitions, types, and function names for FML32 have a "32" suffix (for example, MAXFBLEN32, FBFR32, FLDID32, FLDLEN32, F_OVHD32, Fchg32, and error code Ferror32). Also the environment variables are suffixed with "32" (for example, FLDTBLDIR32, FIELDTBLS32, VIEWFILES32, and VIEWDIR32). For FML32, a fielded buffer pointer is of type "FBFR32 *", a field length has the type FLDLEN32, and the number of occurrences of a field has the type FLDOCC32. Also note that the default required alignment for FML32 buffers is 4-byte alignment.
Existing FML16 applications that are written correctly can easily be changed to use the FML32 interface. All variables used in the calls to the FML functions must use the proper typedefs (FLDID, FLDLEN, and FLDOCC). Any call to tpalloc for an FML typed buffer should use the FMLTYPE definition instead of "FML". The application source code can be changed to use the 32-bit functions simply by changing the include of fml.h to inclusion of fml32.h followed by fml1632.h. The fml1632.h contains macros that convert all of the 16-bit type definitions to 32-bit type definitions, and 16-bit functions and macros to 32-bit functions and macros.
Functions are also provided to convert an FML32 fielded buffer to an FML16 fielded buffer and visa versa.
#include "fml.h" #include "fml32.h" int F32to16(FBFR *dest, FBFR32 *src) int F16to32(FBFR32 *dest, FBFR *src)
F32to16 converts a 32-bit FML buffer to a 16-bit FML buffer. It does this by converting the buffer on a field-by-field basis and then creating the index for the fielded buffer. A field is converted by generating a FLDID from a FLDID32, and copying the field value (and field length for string and carray fields). dest and src are pointers to the destination and source fielded buffers respectively. The source buffer is not changed. These functions can fail for lack of space; they can be re-issued after allocating enough additional space to complete the operation. F16to32 converts a 16-bit FML buffer to a 32-bit FML buffer. It lives in the fml32 library or shared object and sets Ferror32 on error. F32to16 lives in the fml library or shared object and sets Ferror on error. Note that both fml.h and fml32.h must be included to use these functions; fml1632.h may not be included in the same file.
For the remainder of this chapter, rather than continuing to
give both the FML and FML32 names, and VIEW and VIEW32 names, the
16-bit functions will be described.
To make it easier to remember the parameters for the FML functions, a convention has been adopted for the sequence of function parameters. FML parameters appear in the following sequence:
Several functions allow the programmer to query field tables
or given field identifiers for information about fields during
program execution.
Fldid returns the field identifier for a given valid field name and loads the field name/fieldid mapping tables from the field table files, if they do not already exist:
FLDID Fldid(char *name)
where name is a valid field name.
The space used by the mapping tables in memory can be freed
using the Fnmid_unload or Fnmid_unload32
function. Note that these tables are separate from the tables
loaded and used by the Fname function.
Fname returns the field name for a given valid field identifier and loads the fieldid/name mapping tables from the field table files, if they do not already exist:
char * Fname(FLDID fieldid)
where fieldid is a valid field identifier.
The space used by the mapping tables in memory can be freed
using the Fidnm_unload or Fidnm_unload32
function. Note that these tables are separate from the tables
loaded and used by the Fldid function.
Fldno extracts the field number from a given field identifier:
FLDOCC Fldno(FLDID fieldid)
where fieldid is a valid field identifier.
Fldtype extracts the field type (an integer, as defined in fml.h) from a given field identifier.
int Fldtype(FLDID fieldid)
where fieldid is a valid field identifier.
The table in the Figure 1 shows the possible values returned by Fldtype and their meanings.
| Return Value | Meaning |
|---|---|
| 0 | short integer |
| 1 | long integer |
| 2 | character |
| 3 | single-precision float |
| 4 | double-precision float |
| 5 | null-terminated string |
| 6 | character array |
Ftype returns a pointer to a string containing the name of the type of a field given a field identifier:
char * Ftype(FLDID fieldid)
where fieldid is a valid field identifier.
For example:
char *typename . . . typename = Ftype(fieldid);
returns a pointer to one of the following strings: short,
long, char, float, double,
string, or carray.
As part of an application generator, or to reconstruct a field identifier, it might be useful to be able to make a field identifier from a type specification and an available field number. Fmkfldid provides this functionality:
FLDID Fmkfldid(int type, FLDID num)
where
Most FML functions require a pointer to a fielded buffer as an argument. The typedef FBFR is available for declaring such pointers, as in this example:
FBFR *fbfr;
In this chapter, the variable fbfr will be used to mean a pointer to a fielded buffer.
Never attempt to declare fielded buffers themselves, only
pointers to them. The functions used to reserve space for fielded
buffers are explained in the in the following pages, but first we
will describe a function that can be used to determine whether a
given buffer is in fact a fielded buffer.
Fielded (or Fielded32) is used to test whether the specified buffer is fielded.
int Fielded(FBFR *fbfr)
Fielded32 is used with 32-bit FML.
Fielded returns true (1) if the buffer is fielded.
It returns false (0) if the buffer is not fielded and does not
set Ferror in this case.
The amount of memory to allocate for a fielded buffer depends on the maximum number of fields it will contain and the total amount of space needed for all the field values. The function Fneeded can be used to determine the amount of space in bytes needed for a fielded buffer; it takes the number of fields and the space needed for all field values (in bytes) as arguments.
TH Fneeded(FLDOCC F, FLDLEN V)
where
The space needed for field values is computed by estimating the amount of space that would be required by each field value if stored in standard structures (e.g., a long is stored as a long and needs four bytes, etc.). For variable length fields, you should estimate the average amount of space needed for the field. The space calculated by Fneeded includes a fixed overhead for each field; it adds that to the space needed for the field values.
Once you obtain the estimate of space from Fneeded, you can allocate the desired number of bytes using malloc(3) and set up a pointer to the allocated memory space.
For example, the following allocates space for a fielded buffer large enough to contain 25 fields and 300 bytes of values:
#define NF 25
#define NV 300
extern char *malloc;
. . .
if((fbfr = (FBFR *)malloc(Fneeded(NF, NV))) == NULL)
F_error("pgm_name"); /* no space to allocate buffer */
However, this allocated memory space is not yet a fielded
buffer. Finit must be used to initialize it.
The Finit function initializes an allocated memory space as a fielded buffer.
int Finit(FBFR *fbfr, FLDLEN buflen)
where
A call to Finit to initialize the memory space allocated in the previous example above would look like the following:
Finit(fbfr, Fneeded(NF, NV));
Now fbfr points to an initialized, empty fielded buffer. Up to Fneeded(NF, NV) bytes minus a small amount ( F_OVHD as defined in fml.h) are available in the buffer to hold fields.
Note: The numbers used in the malloc(3) (from the previous section) and Finit calls must be the same.
Calls to Fneeded, malloc(3) and Finit may be replaced by a single call to Falloc, which allocates the desired amount of space and initializes the buffer.
FBFR * Falloc(FLDOCC F, FLDLEN V)
where
A call to Falloc that would replace the examples above would look like the following:
extern FBFR *Falloc;
. . .
if((fbfr = Falloc(NF, NV)) == NULL)
F_error("pgm_name"); /* couldn't allocate buffer */
Storage allocated with Falloc (or Fneeded, malloc(3) and Finit) should be freed with Ffree.
Remember that when using System/T, the ATMI functions tpalloc,
tprealloc, and tpfree must be used to
allocate and free message buffers, rather than the FML functions Falloc,
Frealloc, and Free.
Ffree is used to free memory space allocated as a fielded buffer.
int Ffree(FBFR *fbfr)
where
For example:
#include <fml.h>
. . .
if(Ffree(fbfr) < 0)
F_error("pgm_name"); /* not fielded buffer */
Ffree is recommended as opposed to free(3), because Ffree will invalidate a fielded buffer whereas free(3) will not. It is necessary to invalidate fielded buffers because malloc(3) re-uses memory that has been freed, without clearing it. Thus, if free(3) were used, it would be possible for malloc to return a piece of memory that looks like a valid fielded buffer, but is not.
Space for a fielded buffer may also be reserved directly. The buffer must begin on a short boundary. The user must allocate at least F_OVHD bytes (defined in fml.h) for the buffer or an error will be returned from Finit.
The following code is analogous to the preceding example but Fneeded cannot be used to size the static buffer since it is not a macro.
/* the first line aligns the buffer */ static short buffer[500/sizeof(short)]; FBFR *fbfr=(FBFR *)buffer; . . . Finit(fbfr, 500);
It should be emphasized that the following code is quite wrong:
FBFR badfbfr; . . . Finit(&badfbfr, Fneeded(NF, NV));
The structure for FBFR is not defined in the user header files
so this will result in a compilation error.
Fsizeof returns the size of a fielded buffer in bytes:
long Fsizeof(FBFR *fbfr)
where
For example:
long bytes; bytes = Fsizeof(fbfr);
Fsizeof returns the same number that Fneeded
returned when the fielded buffer was originally allocated.
Funused may be used to determine how much space is available in a fielded buffer for additional data:
long Funused(FBFR *fbfr)
where
For example:
long unused; . . . unused = Funused(fbfr);
Note that Funused does not indicate where in the buffer the
unused bytes are, only the number of unused bytes.
Fused may be used to determine how much space is used in a fielded buffer for data and overhead:
long Fused(FBFR *fbfr)
where
For example:
long used; . . . used = Fused(fbfr);
Note that Fused does not indicate where in the buffer the used
bytes are, only the number of used bytes.
At some point, the buffer may run out of space, such as during the addition of a new field value. Frealloc can be used to increase (or decrease) the size of the buffer:
FBFR * Frealloc(FBFR *fbfr, FLDOCC nf, FLDLEN nv)
where
For example:
FBFR *newfbfr;
. . .
if((newfbfr = Frealloc(fbfr, NF+5, NV+300)) == NULL)
F_error("pgm_name"); /* couldn't re-allocate space */
else
fbfr = newfbfr; /* assign new pointer to old */
In this case, the application needed to remember the number of fields and number of value space bytes previously allocated. Note that the arguments to Frealloc (as with its counterpart realloc(3)) are absolute values, not increments. This example will not work if space needs to be re-allocated several times.
The following example shows a second way of incrementing the allocated space:
/* define the increment size when buffer out of space */
#define INCR 400
FBFR *newfbfr;
. . .
if((newfbfr = Frealloc(fbfr, 0, Fsizeof(fbfr)+INCR)) == NULL)
F_error("pgm_name"); /* couldn't re-allocate space */
else
fbfr = newfbfr; /* assign new pointer to old */
Note that you do not need to know the number of fields or the value space size with which the buffer was last initialized. Thus, the easiest way to increase the size is to use the current size plus the increment as the value space. The above example could be executed as many times as needed without remembering past executions or values. The user need not call Finit after calling Frealloc.
If the amount of additional space requested in the call to Frealloc is contiguous to the old buffer, newfbfr and fbfr in the examples above will be the same. However, defensive programming dictates that the user should declare newfbfr as a safeguard against the case where either a new value or NULL is returned. If Frealloc fails, do not use fbfr again.
Note: The buffer size can only be decreased to the number of bytes currently being used in the buffer.
The only restriction on the location of fielded buffers is
that they must be aligned on a short boundary.
Otherwise, fielded buffers are position-independent and may be
moved freely around in memory.
If src points to a fielded buffer and dest points to an area of storage big enough to hold it, then the following might be used to move the fielded buffer:
FBFR *src; char *dest; . . . memcpy(dest, src, Fsizeof(src));
The function memcpy, part of the C runtime memory management functions, moves the number of bytes indicated by its third argument from the area pointed to by its second argument to the area pointed to by its first argument.
While memcpy may be used to copy a fielded buffer, the destination copy of the buffer looks just like the source copy. In particular, for example, the destination copy has the same number of unused bytes as the source buffer.
Fmove acts like memcpy, but does not need an explicit length (it is computed):
int Fmove(char *dest, FBFR *src)
where
For example:
FBFR *src;
char *dest;
. . .
if(Fmove(dest,src) < 0)
F_error("pgm_name");
Fmove checks that the source buffer is indeed a fielded buffer, but does not modify the source buffer in any way.
The destination buffer need not be a fielded buffer (that is,
it need not have been allocated using Falloc), but it
must be aligned on a short boundary (4-byte alignment
for FML32). Thus, Fmove provides an alternative to Fcpy
(see below) when it is desired to copy a fielded buffer to a
non-fielded buffer, but Fmove does not check to make
sure there is enough room in the destination buffer to receive
the source buffer.
Fcpy is used to overwrite one fielded buffer with another:
int Fcpy(FBFR *dest, FBFR *src)
where
Fcpy preserves the overall buffer length of the overwritten fielded buffer; thus, Fcpy is useful for expanding or reducing the size of a fielded buffer. For example:
FBFR *src, *dest;
. . .
if(Fcpy(dest, src) < 0)
F_error("pgm_name");
Unlike Fmove, where dest could point to an uninitialized area, Fcpy expects dest to point to an initialized fielded buffer (allocated using Falloc) and also checks to see that it is big enough to accommodate the data from the source buffer.
Note: You cannot reduce the size of a fielded buffer below the amount of space needed for currently held data.
As with Fmove, the source buffer is not modified by
Fcpy.
This section discusses how to update and access fielded
buffers using the field types of the fields without doing any
conversions. The functions that allow you to convert data from
one type to another upon transfer to/from a fielded buffer are
listed under "Functions That Do Conversion," in this
chapter.
The Fadd function adds a new field value to the fielded buffer.
int Fadd(FBFR *fbfr, FLDID fieldid, char *value, FLDLEN len)
where
If no occurrence of the field exists in the buffer, then the field is added. If one or more occurrences of the field already exist, then the value is added as a new occurrence of the field, and is assigned an occurrence number 1 greater than the current highest occurrence (to add a specific occurrence, Fchg must be used).
Fadd, like all other functions that take or return a field value, expects a pointer to a field value, never the value itself.
If the field type is such that the field length is fixed (short, long, char, float, or double) or can be determined (string), the field length need not be given (it is ignored). If the field type is a character array, the length must be specified; the length is defined as type FLDLEN. For example:
FLDID fieldid, Fldid;
FBFR *fbfr;
. . .
fieldid = Fldid("fieldname");
if(Fadd(fbfr, fieldid, "new value", (FLDLEN)9) < 0)
F_error("pgm_name");
gets the field identifier for the desired field and adds the field value to the buffer.
It is assumed (by default) that the native type of the field is a character array so that the length of the value must be passed to the function. If the value being added is not a character array, the type of value must reflect the type of the value it points to; for instance, the following example adds a long field value:
long lval;
. . .
lval = 123456789;
if(Fadd(fbfr, fieldid, &lval, (FLDLEN)0) < 0)
F_error("pgm_name");
For character array fields, null fields may be indicated by a
length of 0. For string fields, the null string may be stored
since the NULL terminating byte is actually stored as part of the
field value: a string consisting of only the NULL terminating
byte is considered to have a length of 1. For all other types
(fixed length types), you may chose some special value that is
interpreted by the application as a NULL, but the size of the
value will be taken from its field type (e.g., length of four for
a long) regardless of what value is actually passed.
Passing a NULL value address will result in an error, (FEINVAL).
The Fappend function appends a new field value to the fielded buffer.
int Fappend(FBFR *fbfr, FLDID fieldid, char *value, FLDLEN len)
where
Fappend appends a new occurrence of the indicated to the fielded buffer and puts the buffer into append mode. Append mode provides optimized buffer construction for large buffers constructed of many rows of a common set of fields. A buffer that is in append mode is restricted as to what operations may be performed on the buffer. Only calls to the following FML routines are allowed in append mode; Fappend, Findex, Funindex, Ffree, Fused, Funused and Fsizeof. Calls to Findex or Funindex will end append mode. The following example shows the construction of a 500 row buffer with 5 fields per row using Fappend.
for (i=0; i < 500 ;i++) {
if ((Fappend(fbfr, LONGFLD1, &lval1[i], (FLDLEN)0) < 0) ||
(Fappend(fbfr, LONGFLD2, &lval2[i], (FLDLEN)0) < 0) ||
(Fappend(fbfr, STRFLD1, str1[i], (FLDLEN)0) < 0) ||
(Fappend(fbfr, STRFLD2, str2[i], (FLDLEN)0) < 0) ||
(Fappend(fbfr, LONGFLD3, &lval3[i], (FLDLEN)0) < 0)) {
F_error("pgm_name");
break;
}
}
Findex(fbfr, 0);
Fappend, like all other functions that take or return a field value, expects a pointer to a field value, never the value itself.
If the field type is such that the field length is fixed (short, long, char, float, or double) or can be determined (string), the field length need not be given (it is ignored). If the field type is a character array, the length must be specified; the length is defined as type FLDLEN.
It is assumed (by default) that the native type of the field is a character array so that the length of the value must be passed to the function. If the value being appended is not a character array, the type of value must reflect the type of the value it points to.
For character array fields, null fields may be indicated by a
length of 0. For string fields, the null string may be stored
since the NULL terminating byte is actually stored as part of the
field value: a string consisting of only the NULL terminating
byte is considered to have a length of 1. For all other types
(fixed length types), you may chose some special value that is
interpreted by the application as a NULL, but the size of the
value will be taken from its field type (e.g., length of four for
a long) regardless of what value is actually passed.
Passing a NULL value address will result in an error, (FEINVAL).
Fchg changes the value of a field in the buffer.
int Fchg(FBFR *fbfr, FLDID fieldid, FLDOCC oc, char *value, FLDLEN len)
where
For example, to change a field of type carray to a new value stored in value:
FBFR *fbfr;
FLDID fieldid;
FLDOCC oc;
FLDLEN len;
char value[50];
.
.
.
strcpy(value, "new value");
flen = strlen(value);
if(Fchg(fbfr, fieldid, oc, value, len) < 0)
F_error("pgm_name");
If oc is -1, then the field value is added as a new occurrence to the buffer. If oc is 0 or greater and the field is found, then the field value is modified to the new value specified. If oc is 0 or greater and the field is not found, then NULL occurrences are added to the buffer until the value can be added as the specified occurrence. For example, changing field occurrence 3 for a field that does not exist on a buffer will cause three NULL occurrences to be added (occurrences 0, 1 and 2), followed by occurrence 3 with the specified field value. Null values consist of the NULL string "\0" (1 byte in length) for string and character values, 0 for long and short fields, 0.0 for float and double values, and a zero-length string for a character array.
The new or modified value is contained in value. If it is a character array, its length is given in len (len is ignored for other field types). If the value pointer is NULL and the field is found, then the field is deleted. If the field occurrence to be deleted is not found, it is considered an error (FNOTPRES).
The buffer must have enough room to contain the modified or
added field value, or an error is returned (FNOSPACE).
Fcmp compares the field identifiers and field values of two fielded buffers.
int Fcmp( FBFR *fbfr1, FBFR *fbfr2)
where
The function returns a 0 if the buffers are identical; it returns a -1 on any of the following conditions:
Fcmp returns a 1 if any of the reverse set of the
above conditions is true (for example, if the field id of a fbfr2
field is less than the field id of the corresponding field of fbfr1,
etc.).
The Fdel function deletes the specified field occurrence.
int Fdel(FBFR *fbfr, FLDID fieldid, FLDOCC oc)
where
For example,
FLDOCC occurrence;
. . .
occurrence=0;
if(Fdel(fbfr, fieldid, occurrence) < 0)
F_error("pgm_name");
deletes the first occurrence of the field indicated by the
specified field identifier. If it does not exist, the function
returns -1 (Ferror is set to FNOTPRES).
Fdelall deletes all occurrences of the specified field from the buffer:
int Fdelall(FBFR *fbfr, FLDID fieldid)
where
For example:
if(Fdelall(fbfr, fieldid) < 0)
F_error("pgm_name"); /* field not present */
If the field is not found, the function returns -1
(Ferror is set to FNOTPRES).
Fdelete deletes all occurrences of all fields listed in the array of field identifiers, fieldid[]:
int Fdelete(FBFR *fbfr, FLDID *fieldid)
where
The update is done directly to the fielded buffer. The array of field identifiers does not need to be in any specific order, but the last entry in the array must be field identifier 0 (BADFLDID). For example:
#include "fldtbl.h"
FBFR *dest;
FLDID fieldid[20];
.
.
.
fieldid[0] = A; /* field id for field A */
fieldid[1] = D; /* field id for field D */
fieldid[2] = BADFLDID; /* sentinel value */
if(Fdelete(dest, fieldid) < 0)
F_error("pgm_name");
If the destination buffer has fields A, B, C, and D, this example will result in a buffer that contains only occurrences of fields B and C.
Fdelete is a more efficient way of deleting several
fields from a buffer than using several Fdelall calls.
Ffind finds the value of the specified field occurrence in the buffer:
char * Ffind(FBFR *fbfr, FLDID fieldid, FLDOCC oc, FLDLEN *len)
where
In the declaration above, Ffind is shown as returning a char *.
An example of the use of the function is:
#include "fldtbl.h"
FBFR *fbfr;
FLDLEN len;
char *Ffind, *value;
.
.
.
if((value=Ffind(fbfr,ZIP,0, &len)) == NULL)
F_error("pgm_name");
If the field is found, its length is returned in len (if len is NULL, the length is not returned), and its location is returned as the value of the function. If the field is not found, NULL is returned, and Ferror is set to FNOTPRES.
Ffind is useful for gaining "read-only" access to a field. The value returned by Ffind should not be used to modify the buffer. Field value modification should only be done by the functions Fadd or Fchg.
The value returned by Ffind is valid only so long as the buffer remains unmodified. The value is guaranteed to be aligned on a short boundary but may not be aligned on a long or double boundary, even if the field is of that type (see the conversion functions described later in this document for aligned values). On processors that require proper alignment of variables, referencing the value when not aligned properly will cause a system error, as in the following example:
long *l1,l2;
FLDLEN length;
char *Ffind;
.
.
.
if((l1=(long *)Ffind(fbfr, ZIP, 0, &length)) == NULL)
F_error("pgm_name");
else
l2 = *l1;
and should be re-written as:
if((l1==(long *)Ffind(fbfr, ZIP, 0, &length)) == NULL)
F_error("pgm_name");
else
memcpy(&l2,l1,sizeof(long));
This function finds the last occurrence of a field in a fielded buffer and returns a pointer to the field, as well as the occurrence number and length of the field occurrence:
char * Ffindlast(FBFR *fbfr, FLDID fieldid, FLDOCC *oc, FLDLEN *len)
where
In the declaration above, Ffindlast is shown as returning a char *. The actual type of the pointer returned is the same as the type of the value it points to.
Ffindlast acts like Ffind, except that you do not specify a field occurrence. Instead, both the occurrence number and the value of the last field occurrence are returned. However, if you specify NULL for occurrence on calling the function, the occurrence number will not be returned.
The value returned by Ffindlast is valid only as long as the
buffer remains unchanged.
Ffindocc looks at occurrences of the specified field on the buffer and returns the occurrence number of the first field occurrence that matches the user-specified field value:
FLDOCC Ffindocc(FBFR *fbfr, FLDID fieldid, char *value, FLDLEN len;)
where
For example,
#include "fldtbl.h"
FBFR *fbfr;
FLDOCC oc;
long zipvalue;
.
.
.
zipvalue = 123456;
if((oc=Ffindocc(fbfr,ZIP,&zipvalue, 0)) < 0)
F_error("pgm_name");
would set oc to the occurrence for the specified zip code.
Regular expressions are supported for string fields. For example,
#include "fldtbl.h"
FBFR *fbfr;
FLDOCC oc;
char *name;
.
.
.
name = "J.*"
if ((oc = Ffindocc(fbfr, NAME, name, 1)) < 0)
F_error("pgm_name");
would set oc to the occurrence of NAME that starts with 'J'.
Note: To enable pattern matching on strings, the fourth argument to Ffindocc must be nonzero. If zero, simple string compare is performed. If the field value is not found, -1 is returned.
For upward compatibility, a circumflex ( ^ ) and dollar sign ($) are implied to surround the regular expression; thus, the above example is actually interpreted as `^(J.*)$'. This means that the regular expression must match the entire string value in the field.
Fget should be used to retrieve a field from a fielded buffer when the value is to be modified:
int Fget(FBFR *fbfr, FLDID fieldid, FLDOCC oc, char *loc, FLDLEN *maxlen)
where
The caller provides Fget with a pointer to a private buffer, as well as the length of the buffer. If maxlen is specified as NULL, then it is assumed that the destination buffer is large enough to accommodate the field value, and its length is not returned.
Fget returns an error if the desired field is not in the buffer (FNOTPRES), or if the destination buffer is too small (FNOSPACE). For example,
FLDLEN len;
char value[100];
.
.
.
len=sizeof(value);
if(Fget(fbfr, ZIP, 0, value, &len) < 0)
F_error("pgm_name");
gets the zip code assuming it is stored as a character array or string. If it is stored as a long, then it would be retrieved by:
FLDLEN len;
long value;
. . .
len = sizeof(value);
if(Fget(fbfr, ZIP, 0, value, &len) < 0)
F_error("pgm_name");
Like Fget, Fgetalloc finds and makes a copy of a buffer field, but it acquires space for the field via a call to malloc(3):
char * Fgetalloc(FBFR *fbfr, FLDID fieldid, FLDOCC oc, FLDLEN *extralen)
where
In the declaration above, Fgetalloc is shown as returning a char *. The actual type of the pointer returned is the same as the type of the value it points to.
On success, Fgetalloc returns a valid pointer to the copy of the properly aligned buffer field; on error it returns NULL. If malloc(3) fails, Fgetalloc returns an error (Ferror is set to FMALLOC).
The last parameter to Fgetalloc specifies an extra amount of space to be acquired if, for instance, the gotten value is to be expanded before re-insertion into the fielded buffer. On success, the length of the allocated buffer is returned in extralen. For example:
FLDLEN extralen;
FBFR *fieldbfr
char *Fgetalloc;
. . .
extralen = 0;
if (fieldbfr = (FBFR *)Fgetalloc(fbfr, ZIP, 0, &extralen) == NULL)
F_error("pgm_name");
It is the responsibility of the caller to free
space acquired by Fgetalloc.
Fgetlast is used to retrieve the last occurrence of a field from a fielded buffer when the value is to be modified:
int Fgetlast(FBFR *fbfr, FLDID fieldid, FLDOCC *oc, char *loc, FLDLEN *maxlen)
where
The caller provides Fgetlast with a pointer to a private
buffer, as well as the length of the buffer. Fgetlast acts like
Fget, except that you do not specify a field occurrence. Instead,
both the occurrence number and the value of the last field
occurrence are returned. However, if you specify NULL for occ
on calling the function, the occurrence number will not be
returned.
Fnext finds the next field in the buffer after the specified field occurrence:
int Fnext(FBFR *fbfr, FLDID *fieldid, FLDOCC *oc, char *value, FLDLEN *len)
where
A fieldid of FIRSTFLDID should be specified to get the first field in a buffer; the field identifier and occurrence number of the first field occurrence are returned in the corresponding parameters; if the field is not NULL, its value is copied into the memory location addressed by the value pointer; the len parameter is used to determine if value has enough space allocated to contain the field value (Ferror is set to FNOSPACE if it does not); and, the length of the value is returned in the len parameter. Note that if the value of the field is non-null, then the len parameter is also assumed to contain the length of the currently allocated space for value.
If the field value is NULL, then the value and length parameters are not changed.
If no more fields are found, Fnext returns 0 (end of buffer) and fieldid, occurrence, and value are left unchanged.
If the value parameter is not NULL, the length parameter is also assumed to be non-NULL.
The following example reads all field occurrences in the buffer:
FLDID fieldid;
FLDOCC occurrence;
char *value[100];
FLDLEN len;
.
.
.
for(fieldid=FIRSTFLDID,len=sizeof(value);
Fnext(fbfr,&fieldid,&occurrence,value,&len) > 0;
len=sizeof(value)) {
/* code for each field occurrence */
}
Fnum returns the number of fields contained in the specified buffer, or -1 on error:
FLDOCC Fnum(FBFR *fbfr)
where
For example:
if((cnt=Fnum(fbfr)) < 0)
F_error("pgm_name");
else
fprintf(stdout,"%d fields in buffer\n",cnt);
would print the number of fields in the specified buffer.
Foccur returns the number of occurrences for the specified field in the buffer:
FLDOCC Foccur(FBFR *fbfr, FLDID fieldid)
where
Zero is returned if the field does not occur in the buffer and -1 is returned on error. For example:
FLDOCC cnt;
. . .
if((cnt=Foccur(fbfr,ZIP)) < 0)
F_error("pgm_name");
else
fprintf(stdout,"Field ZIP occurs %d times in buffer\n",cnt);
would print the number of occurrences of the field ZIP
in the specified buffer.
Fpres returns true (1) if the specified field occurrence exists and false (0) otherwise:
int Fpres(FBFR *fbfr, FLDID fieldid, FLDOCC oc)
where
For example:
Fpres(fbfr,ZIP,0)
would return true if the field ZIP exists in the
fielded buffer pointed to by fbfr.
Fvals works like Ffind for string values but guarantees that a pointer to a value is returned. Fvall works like Ffind for long and short values, but returns the actual value of the field as a long, instead of a pointer to the value.
| char * | long |
|---|---|
| Fvals(fbfr,fieldid,oc) | Fvall(fbfr,fieldid,oc) |
| FBFR *fbfr; | FBFR *fbfr; |
| FLDID fieldid; | FLDID fieldid; |
| FLDOCC oc; | FLDOCC oc; |
where
For Fvals, if the specified field occurrence is not found, the NULL string, \0, is returned. This function is useful for passing the value of a field to another function without checking the return value. This function is valid only for fields of type string; the NULL string is automatically returned for other field types (i.e., no conversion is done).
For Fvall, if the specified field occurrence is not found,
then 0 is returned. This function is useful for passing the value
of a field to another function without checking the return value.
This function is valid only for fields of type long
and short; 0 is automatically returned for other field
types (i.e., no conversion is done).
The functions listed in this section access and update entire fielded buffers, rather than individual fields in the buffers. These functions use at most three parameters, dest, src, and fieldid, where
Fconcat adds fields from the source buffer to the fields that already exist in the destination buffer.
int Fconcat(FBFR *dest, FBFR *src)
Occurrences in the destination buffer are maintained (i.e., retained and not modified) and new occurrences from the source buffer are added with greater occurrence numbers than any existing occurrences for each field (the fields are maintained in field identifier order).
In the following example:
FBFR *src, *dest;
. . .
if(Fconcat(dest,src) < 0)
F_error("pgm_name");
if dest has fields A, B, and two occurrences of C, and src has fields A, C, and D, the resultant dest will have two occurrences of field A (destination field A and source field A), field B, three occurrences of field C (two from dest and the third from src), and field D.
This operation will fail if there is not enough space to
contain the new fields (FNOSPACE); in this case, the
destination buffer remains unchanged.
Fjoin is used to join two fielded buffers based on matching fieldid/occurrence.
int Fjoin(FBFR *dest, FBFR *src)
For fields that match on fieldid/occurrence, the field value is updated in the destination buffer with the value from the source buffer. Fields in the destination buffer that have no corresponding fieldid/occurrence in the source buffer are deleted. Fields in the source buffer that have no corresponding fieldid/occurrence in the destination buffer are not added to the destination buffer. Thus,
if(Fjoin(dest,src) < 0)
F_error("pgm_name");
Using the input buffers in the previous example will result in
a destination buffer that has source field value A and source
field value C. This function may fail due to lack of space if the
new values are larger than the old (FNOSPACE); in this
case, the destination buffer will have been modified. However, if
this happens, the destination buffer may be re-allocated using
Frealloc and the Fjoin function repeated (even if the destination
buffer has been partially updated, repeating the function will
give the correct results).
Fojoin is similar to Fjoin, but it does not delete fields from the destination buffer that have no corresponding fieldid/occurrence in the source buffer.
int Fojoin(FBFR *dest, FBFR *src)
Note that fields that exist in the source buffer that have no corresponding fieldid/occurrence in the destination buffer are not added to the destination buffer. For example:
if(Fojoin(dest,src) < 0)
F_error("pgm_name");
Using the input buffers from the previous example, dest
will contain the source field value A, the destination field
value B, the source field value C, and the second destination
field value C. As with Fjoin, this function can fail for lack of
space (FNOSPACE) and can be re-issued again after
allocating more space to complete the operation.
Fproj is used to update a buffer in place so that only the desired fields are kept (in other words, the result is a projection on specified fields).
int Fproj(FBFR *fbfr, FLDID *fieldid)
These fields are specified in an array of field identifiers passed to the function. The update is performed directly in the fielded buffer. For example:
#include "fldtbl.h"
FBFR *fbfr;
FLDID fieldid[20];
.
.
.
fieldid[0] = A; /* field id for field A */
fieldid[1] = D; /* field id for field D */
fieldid[2] = BADFLDID; /* sentinel value */
if(Fproj(fbfr, fieldid) < 0)
F_error("pgm_name");
If the buffer has fields A, B, C, and D, the example results
in a buffer that contains only occurrences of fields A and D.
Note that the entries in the array of field identifiers do not
need to be in any specific order, but the last value in the array
of field identifiers must be field identifier 0 (BADFLDID).
Fprojcpy is similar to Fproj but the projection is done into a destination buffer.
int Fprojcpy(FBFR *dest, FBFR *src, FLDID *fieldid)
Any fields in the destination buffer are first deleted and the results of the projection on the source buffer are copied into the destination buffer. Using the above example,
if(Fprojcpy(dest, src, fieldid) < 0)
F_error("pgm_name");
will place the results of the projection in the destination
buffer. The entries in the array of field identifiers may be
re-arranged; the field identifier array is sorted if they are not
in numeric order.
Fupdate updates the destination buffer with the field values in the source buffer.
int Fupdate(FBFR *dest, FBFR *src)
For fields that match on fieldid/occurrence, the field value is updated in the destination buffer with the value in the source buffer (like Fjoin). Fields on the destination buffer that have no corresponding field on the source buffer are left untouched (like Fojoin). Fields on the source buffer that have no corresponding field on the destination buffer are added to the destination buffer (like Fconcat). For example:
if(Fupdate(dest,src) < 0)
F_error("pgm_name");
If the src buffer has fields A, C, and D, and the dest
buffer has fields A, B, and two occurrences of C, the updated
destination buffer will contain: the source field value A, the
destination field value B, the source field value C, the second
destination field value C, and the source field value D.
This function transfers data from a fielded buffer to a C structure using a specified view description.
int Fvftos(FBFR *fbfr, char *cstruct, char *view)
where
If the named view is not found, Fvftos returns -1, and Ferror is set to FBADVIEW.
When transferring data from a fielded buffer to a C structure, the following rules apply:
For example,
#include <stdio.h> #include "fml.h" #include "custdb.flds.h" #include "custdb.h" struct custdb cust; FBFR *fbfr; . . . fbfr = Falloc(800,1000); Fvinit((char *)&cust,"custdb"); /* initialize cust */ str = "string1"; Fadd(fbfr,ACTION,str,(FLDLEN)8); str = "abc"; Fadd(fbfr,BUG_CURS,str,(FLDLEN)4); Fvftos(fbfr,(char *)&cust,"custdb"); . . .
would put "string1" into cust.action[0] and "abc" into cust.bug[0]. All other members in the cust structure should contain null values.
View custdb is defined in ``VIEWS Examples'' in
Chapter 6.
This function transfers data from a C structure to a fielded buffer using a specified view description.
int Fvstof(FBFR *fbfr, char *cstruct, int mode, char *view)
where
The transfer process obeys the rules listed under the FML function corresponding to the mode parameter (Fupdate, Fjoin, Fojoin, and Fconcat, described in this chapter).
If the named view is not found, Fvstof returns -1, and Ferror is set to FBADVIEW.
Note: Null values are not transferred from a structure member to a fielded buffer. That is, during a structure-to-field transfer, if a structure member contains the (default or user-specified) null value defined for that member, the member is ignored.
Fvnull is used to determine if an occurrence in a C structure contains the null value for that field.
int Fvnull(char *cstruct, char *cname, FLDOCC oc, char *view)
where
Fvnull returns:
1 if an occurrence is null 0 if an occurrence is not null -1 if an error occurred
This function initializes all elements in a C structure to their appropriate null value.
int Fvsinit(char *cstruct, char *view)
where
This function allows users to change flag options at run time.
int Fvopt(char *cname, int option, char *view)
where
Possible values for the option parameter are:
Note that changes to view descriptions are not permanent. They
are guaranteed only until another view description is accessed.
This function initializes an individual member of a C structure to its appropriate null value. It sets the ACM of the element to 0, if the C flag is used in the view file; it sets the ALMs to the length of the associated null value, if the L flag is used in the view file.
int Fvselinit(char *cstruct, char *cname, char *view)
where
FML provides a set of routines that perform data conversion upon reading or writing a fielded buffer.
Generally, the functions behave like their non-conversion counterparts, except that they provide conversion from a user type to the native field type when writing to a buffer, and from the native type to a user type when reading from a buffer.
The native type of a field is the type specified for it in its
field table entry and encoded in its field identifier. (the only
exception to this rule is CFfindocc, which, although it is a read
operation, converts from the user-specified type to the native
type before calling Ffindocc). The function names are the same as
their non-conversion FML counterparts with a "C"
prefix.
The CFadd function adds a user supplied item to a buffer creating a new field occurrence within the buffer:
int CFadd(FBFR *fbfr, FLDID fieldid, char *value, FLDLEN len, int type)
where
Before the field addition, the data item is converted from a user supplied type to the type specified in the field table as the fielded buffer storage type of the field. If the source type is FLD_CARRAY (character array), the length argument should be set to the length of the array. For example,
if(CFadd(fbfr,ZIP,"12345",(FLDLEN)0,FLD_STRING) <0)
F_error("pgm_name");
If the ZIP (zip code) field were stored in a fielded buffer as a long integer, the function would convert "12345" to a long integer representation, before adding it to the fielded buffer pointed to by fbfr (note that the field value length is given as 0 since the function can determine it; the length is needed only for type FLD_CARRAY). The following code fragment:
long zipval;
. . .
zipval = 12345;
if(CFadd(fbfr,ZIP,&zipval,(FLDLEN)0,FLD_LONG) < 0)
F_error("pgm_name");
puts the same value into the fielded buffer, but does so by
presenting it as a long, instead of as a string. Note that the
value must first be put into a variable, since C does not permit
the construct &12345L. CFadd returns 1 on success,
and -1 on error, in which case Ferror is
set appropriately.
The function CFchg acts like CFadd, except that it changes the value of a field (after conversion of the supplied value):
int CFchg(FBFR *fbfr, FLDID fieldid, FLDOCC oc, char *value, FLDLEN len, int type)
where
For example,
FLDOCC occurrence;
long zipval;
. . .
zipval = 12345;
occurrence = 0;
if(CFchg(fbfr,ZIP,occurrence,&zipval,(FLDLEN)0,FLD_LONG) < 0)
F_error("pgm_name");
would change the first occurrence (occurrence 0) of field ZIP to the specified value, doing any conversion if needed.
If the specified occurrence is not found, then null
occurrences are added to pad the buffer with multiple occurrences
until the value can be added as the specified occurrence.
CFget is the conversion analog of Fget. The difference is that it copies a converted value to the user-supplied buffer:
int CFget(FBFR *fbfr, FLDID fieldid, FLDOCC oc, char *buf, FLDLEN *len, int type)
where
Using the previous example,
FLDLEN len;
. . .
len=sizeof(zipval);
if(CFget(fbfr,ZIP,occurrence,&zipval,&len,FLD_LONG) < 0)
F_error("pgm_name");
would get the value that was just stored in the buffer, no
matter what format, and convert it back to a long integer. If the
length pointer is NULL, then the length of the value retrieved
and converted is not returned.
CFgetalloc, is like Fgetalloc; you are responsible for freeing the malloc'd space for the returned (converted) value with free:
char * CFgetalloc(FBFR *fbfr, FLDID fieldid, FLDOCC oc, int type, FLDLEN *extralen)
where
In the declaration above, CFgetalloc is shown as returning a char *.
The previously stored value could be retrieved into space allocated automatically for you by the following code:
char *value;
FLDLEN extra;
. . .
extra = 25;
if((value=CFgetalloc(fbfr,ZIP,0,FLD_LONG,&extra)) == NULL)
F_error("pgm_name");
The value extra in the function call indicates that
the function should not only allocate enough space for the
retrieved value but an additional 25 bytes and the total amount
of space allocated will be returned in this variable.
CFfind returns a pointer to a converted value of the desired field:
char * CFfind(FBFR *fbfr. FLDID fieldid, FLDOCC oc, FLDLEN len, int type)
where
In the declaration above, CFfind is shown as returning a char *. The actual type of the pointer returned is the same as the type of the value it points to.
Like Ffind, this pointer should be considered read only. For example:
char *CFfind;
FLDLEN len;
long *value;
. . .
if((value=(long *)CFfind(fbfr,ZIP,occurrence,&len,FLD_LONG))==NULL)
F_error("pgm_name");
would return a pointer to a long containing the value of the first occurrence of the ZIP field. If the length pointer is NULL, then the length of the value found is not returned. Unlike Ffind, the value returned is guaranteed to be properly aligned for the corresponding user-specified type.
Note: The duration of the validity of the pointer returned by CFfind is guaranteed only until the next buffer operation, even if it is non-destructive, since the converted value is retained in a single private buffer. This differs from the value returned by Ffind, which is guaranteed until the next modification of the buffer.
CFfindocc looks at occurrences of the specified field on the buffer and returns the occurrence number of the first field occurrence that matches the user-specified field value after it has been converted (it is converted to the type of the field identifier).
FLDOCC CFfindocc(FBFR *fbfr, FLDID fieldid, char *value, FLDLEN len, int type)
where
For example,
#include "fldtbl.h"
FBFR *fbfr;
FLDOCC oc;
char zipvalue[20];
. . .
strcpy(zipvalue,"123456");
if((oc=CFfindocc(fbfr,ZIP,zipvalue,0,FLD_STRING)) < 0)
F_error("pgm_name");
would convert the string to the type of fieldid ZIP (possibly a long) and set oc to the occurrence for the specified zip code. If the field value is not found, -1 is returned.
Note: Since CFfindocc converts the user-specified value to the native field type before examining the field values, regular expressions will only work when the user-specified type and the native field type are both FLD_STRING. Thus, CFfindocc has no utility with regular expressions.
A set of functions ( Fadds(3fml), Fchgs(3fml), Fgets(3fml), Fgetsa(3fml) and Ffinds(3fml) and their FML32 counterparts) has been provided to handle the case of conversion to/from a user type of FLD_STRING. These functions call their non-string-function counterparts, providing a type of FLD_STRING, and a len of 0. Note that the duration of the validity of the pointer returned by Ffinds is the same as that described for CFfind.
See the BEA TUXEDO
Reference Manual: Section 3fml for descriptions of these
functions.
The functions CFadd, CFchg, CFget, CFgetalloc, and CFfind use the function Ftypcvt to perform the appropriate data conversion. The synopsis of Ftypcvt usage is as follows (it does not follow the parameter order conventions):
char * Ftypcvt(FLDLEN *tolen, int totype, char *fromval, int fromtype, FLDLEN fromlen)
where
Ftypcvt converts from the value *fromval, which has type fromtype, and length fromlen if fromtype is type FLD_CARRAY (otherwise fromlen is inferred from fromtype), to a value of type totype. Ftypcvt returns a pointer to the converted value, and sets *tolen to the converted length, upon success. Upon failure, Ftypcvt returns NULL. As an example of its usage, the function CFchg is presented:
CFchg(fbfr,fieldid,oc,value,len,type)
FBFR *fbfr; /* fielded buffer */
FLDID fieldid; /* field to be changed */
FLDOCC oc; /* occurrence of field to be changed */
char *value; /* location of new value */
FLDLEN len; /* length of new value */
int type; /* type of new value */
{
char *convloc; /* location of post-conversion value */
FLDLEN convlen; /* length of post-conversion value */
extern char *Ftypcvt;
/* convert value to fielded buffer type */
if((convloc = Ftypcvt(&convlen,FLDTYPE(fieldid),value,type,
len)) == NULL)
return(-1);
if(Fchg(fbfr,fieldid,oc,convloc,convlen) < 0)
return(-1);
return(1);
}
The user may call Ftypcvt directly to do field value
conversion without adding or modifying a fielded buffer.
A description of conversion rules is now presented. In this description, oldval represents a pointer to the data item being converted, and newval a pointer to the post-conversion value:
*((float *)newval) = *((short *) oldval)
sprintf(newval,"%d",*((short *)oldval))
*((float *)newval) = atof(oldval)
*((float *)newval) = *((char *)oldval)
is the method used to convert a char to a float. Similarly,
*((char *)newval) = *((short *)oldval)
is used to convert a short to a char.
Note that a carray of length 1 and a char have the following differences:
The tables in Figures 2 and 3 summarize the conversion rules presented in this section:
| dest type | ||||||||
|---|---|---|---|---|---|---|---|---|
| src typ | char | short | long | float | double | string | carray | dec_t |
| char | - | cast | cast | cast | cast | st[0]=c | array[0]=c | d |
| short | cast | - | cast | cast | cast | sprintf | sprintf | d |
| long | cast | cast | - | cast | cast | sprintf | sprintf | d |
| float | cast | cast | cast | - | cast | sprintf | sprintf | d |
| double | cast | cast | cast | cast | - | sprintf | sprintf | d |
| string | c=st[0] | atoi | atol | atof | atof | - | drop 0 | d |
| carray | c=array[0] | atoi | atol | atof | atof | add 0 | - | d |
| dec_t | d | d | d | d | d | d | d | - |
The entries in the above table have the following meanings:
| table entry | meaning |
|---|---|
| - | no conversion need be done, src and dest are same type |
| cast | conversion done using C assignment with type casting |
| sprintf | conversion done using sprintf function |
| atoi | conversion done using atoi function |
| atof | conversion done using atof function |
| atol | conversion done using atol function |
| add 0 | conversion done by concatenating NULL byte |
| drop 0 | conversion done by dropping terminating NULL byte |
| c=array[0] | character set to first byte of array |
| array[0]=c | first byte of array is set to character |
| c=st[0] | character set to first byte of string |
| st[0]=c | first byte of string set to c |
| d | decimal(3) conversion function |
When a fielded buffer is initialized by Finit or Falloc, an index is automatically set up. This index is used to expedite fielded buffer accesses and is transparent to you. As fields are added to or deleted from the fielded buffer, the index is automatically updated.
However, when storing a fielded buffer on a long-term storage
device, or when transferring it between cooperating processes, it
may be desirable to save space by eliminating its index and
regenerating it upon receipt. The functions described in this
section may be used to perform such index manipulations.
This function returns the amount of space used by the index of a buffer:
long Fidxused(FBFR *fbfr)
where
You can use this function to determine the size of the index
of a buffer and whether significant time or space would be saved
by deleting the index.
The function Findex may be used at any time to index an unindexed fielded buffer:
int Findex(FBFR *fbfr. FLDOCC intvl)
where
The second argument to Findex specifies the indexing interval for the buffer. If 0 is specified, the value FSTDXINT (defined in fml.h) is used. The user may ensure that all fields are indexed by specifying an interval of 1.
Note that more space may be made available in an existing
buffer for user data by increasing the indexing interval, and
re-indexing the buffer. This represents a space/time tradeoff,
however, since reducing the number of index elements (by
increasing the index interval), means, in general, that searches
for fields will take longer. Most operations will attempt to drop
the entire index if they run out of space before returning a
"no space" error.
This function can be used instead of Findex in cases where the fielded buffer has not been altered since its index was removed:
int Frstrindex(FBFR *fbfr, FLDOCC numidx)
where
Funindex discards the index of a fielded buffer and returns the number of index entries the buffer had before the index was stripped:
FLDOCC Funindex(FBFR *fbfr)
where
To transmit a fielded buffer without its index, something similar to the following should be done:
save = Funindex(fbfr);
num_to_send = Fused(fbfr);
transmit(fbfr,num_to_send);
Frstrindex(fbfr,save);
On the receiving side, the index could be regenerated with the following statement:
Findex(fbfr);
Note that the receiving process cannot call Frstrindex because it did not remove the index itself, and the index was not sent with the file.
Note: The space used in memory by the index is not freed by calling Funindex; this function only saves space on disk or when sending a buffer to another process. Of course, you are always free to send a fielded buffer and its index to another process and avoid using these functions.
The functions described in this section provide for input and
output of fielded buffers to standard I/O or to file streams.
The I/O functions Fread and Fwrite work with the Standard I/O Library:
int Fread(FBFR *fbfr, FILE *iop) int Fwrite(FBFR *fbfr, FILE *iop)
The stream to or from which the I/O is directed is determined by a FILE pointer argument. This argument must be set up using the normal Standard I/O Library functions.
A fielded buffer may be written into a Standard I/O stream with the function Fwrite, like this:
if (Fwrite(fbfr, iop) < 0)
F_error("pgm_name");
A buffer written with Fwrite may be read with Fread, as in:
if(Fread(fbfr, iop) < 0)
F_error("pgm_name");
Although the contents of the fielded buffer pointed to by fbfr are replaced by the fielded buffer read in, the capacity of the fielded buffer (size of the buffer) remains unchanged.
Fwrite discards the buffer index, writing only as much of the fielded buffer as has been used (as returned by Fused).
Fread restores the index of a buffer by calling Findex. The
buffer is indexed with the same indexing interval with which it
was written by Fwrite.
A checksum may be calculated for verifying I/O:
long chk; . . . chk = Fchksum(fbfr);
The user is responsible for calling Fchksum, writing the
checksum value out along with the fielded buffer, and checking it
on input. Fwrite does not write the checksum automatically.
The function Fprint prints a fielded buffer on the standard output in ASCII format:
Fprint(FBFR *fbfr)
where
Ffprint is similar to Fprint, except the text is printed to a specified output stream:
Ffprint(FBFR *fbfr, FILE *iop)
where
Each of these print functions prints, for each field
occurrence, the field name and the field value, separated by a
tab and followed by a new-line. Fname is used to determine the
field name; if the field name cannot be determined, then the
field identifier is printed. Non-printable characters in the
field values for strings and character arrays are represented by
a backslash followed by their two-character hexadecimal value.
Backslashes occurring in the text are escaped with an extra
backslash. A blank line is printed following the output of the
printed buffer.
Fextread may be used to construct a fielded buffer from its printed format, i.e. from the output of Fprint (hexadecimal values output by Fprint are interpreted properly).
int Fextread(FBFR *fbfr, FILE *iop)
Fextread accepts an optional flag preceding the field-name/ field-identifier specification in the output of Fprint as follows:
| flag | indicates |
|---|---|
| + | field should be changed in the buffer |
| - | field should be deleted from the buffer |
| = | one field should be assigned to another |
| # | comment line - ignored |
If no flag is given, the default action is to Fadd the field to the buffer.
Field values may be extended across lines by having the overflow lines begin with a tab (the tab is discarded). A single blank line signals end of buffer; successive blank lines yield a null buffer.
If an error has occurred, -1 is returned, and Ferror
is set accordingly. If end of file is reached before a blank
line, Ferror is set to FSYNTAX.
The functions in this section deal with the evaluation of boolean expressions where the "variables" of the expression are the values of fields in a fielded buffer or a VIEW. Functions described in this section allow you to:
A function is provided that compiles the expression into a
compact form suitable for efficient evaluation. A second function
evaluates the compiled form against a fielded buffer to produce a
true or false answer.
This section describes, in detail, the expressions accepted by the boolean compilation function and how the expression is evaluated. Figure 4 shows the Backus-Naur Form definitions of the accepted boolean expressions.
Standard C language operators not supported include the shift operators (<< and >>), the bitwise "or" and "and" operators (& and |), the conditional operator (?), the prefix and postfix incrementation operators (++ and --), the address and indirection operators (& and *), the assignment operator (=), and the comma operator (,). The following sections describe boolean expressions in greater detail.
| expression | definition |
|---|---|
| <boolean> | <boolean> || <logical and> | <logical and> |
| <logical and> | <logical and> && <xor expr> | <xor expr> |
| <xor expr> | <xor expr> ^ <equality expr> | <equality expr> |
| <equality expr> | <equality expr> <eq op> <relational expr> | <relational expr> |
| <eq op> | == | != | %% | !% |
| <relational expr> | <relational expr> <rel op> <additive expr> | <additive expr> |
| <rel op> | < | <= | >= | > | |
| <additive expr> | <additive expr> <add op> <multiplicative expr> | <multiplicative expr> |
| <add op> | + | - |
| <multiplicative expr> | <multiplicative expr> <mult op> <unary expr> | <unary expr> |
| <mult op> | * | / | % |
| <unary expr> | <unary op> <primary expr> | <primary expr> |
| <unary op> | + | - | ~ | ! |
| <primary expr> | ( <boolean> ) | <unsigned constant> | <field ref> |
| <unsigned constant> | <unsigned number> | <string> |
| <unsigned number> | <unsigned float> | <unsigned int> |
| <string> | ' <character> {<character>...} ' |
| <field ref> | <field name> | <field name>[<field occurrence>] |
| <field occurrence> | <unsigned int> | <meta> |
| <meta> | ? |
The only variables allowed in the boolean expressions are field references. There are several restrictions on field names. Names are made up of letters and digits; the first character must be a letter. The underscore (_) counts as a letter; it is useful for improving the readability of long variable names. Up to 30 characters are significant. There are no reserved words.
For a fielded buffer evaluation, any field that is referenced in a boolean expression must exist in a field table. This implies that the FLDTBLDIR and FIELDTBLS environment variables are set, as described in Chapter 3, before using the boolean compilation function. The field types used in booleans are those allowed for FML fields; namely, short, long, float, double, char, string, and carray. Along with the field name, the field type is kept in the field table. Thus, the field type can always be determined.
For a VIEW evaluation, any field that is referenced in a
boolean expression must exist as a C structure element name, not
the associated fielded buffer name, in the VIEW. This implies
that the VIEWDIR and VIEWFILES environment
variables are set, as described in Chapter 3, before using the
boolean compilation function. The field types used in booleans
are those allowed for FML VIEWS; namely, short, long, float,
double, char, string, carray, plus int and dec_t. Along with the
field name, the field type is kept in the view definition. Thus,
the field type can always be determined.
A string is a group of characters within single quotes. The ASCII code for a character may be substituted for the character via an escape sequence. An escape sequence takes the form of a backslash followed by exactly two hexadecimal digits (NOTE THAT THIS IS NOT AS IN C where a hexadecimal escape sequence starts with \x).
As an example, consider 'hello' and 'hell\\6f'. They are equivalent strings because the hexadecimal code for an 'o' is 6f.
Octal escape sequences and escape sequences such as
"\n" are not supported.
Numeric integer and floating point constants are accepted, as
in C (octal and hexadecimal constants are not recognized).
Integer constants are treated as longs and floating point
constants are treated as doubles (decimal constants for the dec_t
type are not supported).
To evaluate a boolean expression, the following conversions are performed by the boolean compiler:
Boolean expressions are built from primary expressions, which can be any of the following:
A field name or a field name followed by a subscript is a primary expression. The subscript indicates which occurrence of the field is being referenced. The subscript may be either an integer constant, or '?' indicating any occurrence; the subscript cannot be an expression. If the field name is not subscripted, field occurrence 0 is assumed.
If a field name reference appears without an arithmetic, unary, equality, or relational operator, then its value is the long integer value 1 if the field exists and 0 if the field does not exist. This may be used to test the existence of a field in the fielded buffer regardless of field type (note that there is no * indirection operator).
A constant is a primary expression. Its type may be long, double, or carray, as discussed in the conversion section.
A parenthesized expression is a primary expression whose type
and value are identical to those of the unadorned expression.
Parentheses may be used to change the precedence of operators,
which is discussed in the next section.
The table in Figure 5 lists the precedence of expression operators, with the operators having the highest precedence at the top of the list:
| type | operators |
|---|---|
| unary | +, -, !, ~ |
| multiplicative | *, /, % |
| additive | +, - |
| relational | <, >, <=, >=, ==, != |
| equality and matching | ==, !=, %%, !% |
| exclusive OR | ^ |
| logical AND | && |
| logical OR | || |
Within each operator type, the operators have the same
precedence. The following sections discuss each operator type in
detail. As in C, you can override the precedence of operators by
using parentheses.
The unary operators recognized are the unary plus operator (+), the unary minus operator (-), the one's complement operator (~), and the logical not operator (!). Expressions with unary operators group right-to-left:
+ expression - expression ~ expression ! expression
The unary plus operator has no effect on the operand (it is recognized and ignored). The result of the unary minus operator is the negative of its operand. The usual arithmetic conversions are performed. Unsigned entities do not exist in FML and thus cause no problems with this operator.
The result of the logical negation operator is 1 if the value of its operand is 0, and 0 if the value of its operand is non-zero. The type of the result is long.
The result of the one's complement operator is the one's
complement of its operand. The type of the result is long.
The multiplicative operators *, /, and % group left-to-right. The usual arithmetic conversions are performed.
expression * expression expression / expression expression % expression
The binary * operator indicates multiplication. The * operator is associative and expressions with several multiplications at the same level may be rearranged by the compiler.
The binary / operator indicates division. When positive integers are divided truncation is toward 0, but the form of truncation is machine-dependent if either operand is negative.
The binary % operator yields the remainder from the division
of the first expression by the second. The usual arithmetic
conversions are performed. The operands must not be float or
double.
The additive operators + and - group left-to-right. The usual arithmetic conversions are performed.
expression + expression expression - expression
The result of the + operator is the sum of the operands. The operator + is associative and expressions with several additions at the same level may be rearranged by the compiler. The operands must not both be strings; if one is a string, it is converted to the arithmetic type of the other.
The result of the - operator is the difference of the
operands. The usual arithmetic conversions are performed. The
operands must not both be strings; if one is a string, it is
converted to the arithmetic type of the other.
These operators group left-to-right.
expression == expression expression != expression expression %% expression expression !% expression
The == (equal to) and the != (not equal to) operators yield 0 if the specified relation is false and 1 if it is true. The type of the result is long. The usual arithmetic conversions are performed.
The %% operator takes, as its second expression, a regular expression against which it matches its first expression. The second expression (the regular expression) must be a quoted string. The first expression may be a FML field name, or a quoted string. This operator yields a 1 if the first expression is fully matched by the second expression (the regular expression). The operator yields a 0 in all other cases.
The !% operator is the not regular expression match operator. It takes exactly the same operands as the %% operator, but yields exactly the opposite results. The relationship between %% and !% is analogous to the relationship between == and !=.
The regular expressions allowed are described on the recomp(3c) manual
page.
These operators group left-to-right.
expression < expression expression > expression expression <= expression expression >= expression
The operators < (less than), > (greater than), <=
(less than or equal to) and >= (greater than or equal to) all
yield 0 if the specified relation is false and 1 if it is true.
The type of the result is long. The usual arithmetic conversions
are performed.
The ^ operator groups left-to-right.
expression ^ expression
It returns the bitwise exclusive OR function of the operands.
The result is always a long.
expression && expression
The && operator groups left-to-right. It returns 1 if
both its operands are non-zero, 0 otherwise. The &&
operator guarantees left-to-right evaluation. However, it is not
guaranteed that the second operand is not evaluated if the first
operand is 0; this is different from the C language. The operands
need not have the same type. The result is always a long.
The || operator groups left-to-right.
expression || expression
It returns 1 if either of its operands is non-zero, and 0
otherwise. The || operator guarantees left-to-right
evaluation. However, it is not guaranteed that the second operand
is not evaluated if the first operand is non-zero; this is
different from the C language. The operands need not have the
same type, and the result is always a long.
The following field table defines the fields used for the sample boolean expressions:
EMPID 200 carray SEX 201 char AGE 202 short DEPT 203 long SALARY 204 float NAME 205 string
Recall that boolean expressions always evaluate to either true or false. The following example:
"EMPID[2] %% '123.*' && AGE < 32"
would be true if field occurrence 2 of EMPID exists and begins with the characters "123" and the age field (occurrence 0) appears and is less than 32. This example uses a constant integer as a subscript to EMPID. The ? subscript is used in the following example:
"PETS[?] == 'dog'"
This expression would be true if PETS exists and
any occurrence of it contained the characters "dog".
The following sections describe the various functions that
take boolean expressions as arguments.
Fboolco compiles a boolean expression for FML and returns a pointer to an evaluation tree:
char * Fboolco(char *expression)
where *expression is a pointer to an expression to be compiled.
Fvboolco compiles a boolean expression for a VIEW and returns a pointer to an evaluation tree:
char * Fvboolco(char *expression, char *viewname)
where *expression is a pointer to an expression to be compiled, and *viewname is a pointer to the view name for which the fields are evaluated.
Space is allocated using malloc(3) to hold the evaluation tree. For example,
#include <stdio.h>
#include "fml.h"
extern char *Fboolco;
char *tree;
. . .
if((tree=Fboolco("FIRSTNAME %% 'J.*n' && SEX == 'M'")) == NULL)
F_error("pgm_name");
would compile a boolean expression that checks if the FIRSTNAME field is in the buffer, begins with 'J' and ends with 'n' (e.g., John, Joan, etc.) and the SEX field equal to 'M'.
The first and second characters of the tree array form the least significant byte and the most significant byte, respectively, of an unsigned 16 bit quantity that gives the length, in bytes, of the entire array. This value is useful for copying or otherwise manipulating the array.
The evaluation tree produced by Fboolco is used by the other boolean functions listed below; this avoids having to constantly re-compile the expression.
free(3) should be used to free the space allocated
to an evaluation tree when the boolean expression will no longer
be used. Compiling many boolean expressions without freeing the
evaluation tree when no longer needed may cause a program to run
out of data space. free(3)
Fboolpr prints a compiled expression to the specified file stream. The expression is fully parenthesized, as it was parsed (as indicated by the evaluation tree),
void Fboolpr(char *tree, FILE *iop)
where
Fvboolpr prints a compiled expression to the specified file stream.
void Fvboolpr(char *tree, FILE *iop, char *viewname)
where
This function is useful for debugging.
Executing Fboolpr on the expression compiled above would yield
(((FIRSTNAME[0]) %% ('J.*n')) && ((SEX[0]) == ('M')))
These functions evaluate a fielded buffer against a boolean expression against a fielded buffer:
| int | double |
|---|---|
| Fboolev(fbfr,tree) | Ffloatev(fbfr,tree) |
| FBFR *fbfr; | FBFR *fbfr; |
| char *tree; | char *tree; |
where
The VIEW equivalents are as follows.
| int | double |
|---|---|
| Fvboolev(fbfr,tree,viewname) | Fvfloatev(fbfr,tree,viewname) |
| FBFR *fbfr; | FBFR *fbfr; |
| char *tree; | char *tree; |
| char *viewname; | char *viewname; |
Fboolev returns true (1) if the fielded buffer matches the boolean conditions specified in the evaluation tree. This function does not change either the fielded buffer or the evaluation tree. Using the evaluation tree compiled above:
#include <stdio.h> #include "fml.h" #include "fldtbl.h" FBFR *fbfr; . . . Fchg(fbfr,FIRSTNAME,0,"John",0); Fchg(fbfr,SEX,0,"M",0); if(Fboolev(fbfr,tree) > 0) fprintf(stderr,"Buffer selected\n"); else fprintf(stderr,"Buffer not selected\n");
would print "Buffer selected".
Ffloatev or Ffloatev32 is similar to Fboolev, but returns the value of the expression as a double. For example,
#include <stdio.h>
#include "fml.h"
. . .
FBFR *fbfr;
main {
char *Fboolco;
char *tree;
double Ffloatev;
if (tree=Fboolco("3.3+3.3")) {
printf("%lf",Ffloatev(fbfr,tree));
}
}
would print 6.6. If Fboolev were used in place of Ffloatev
in the above example, a 1 would be printed.
A VIEW can be converted to and from a target record format.
The default target format is IBM System/370 COBOL records.
The three functions that provide target conversion are as follows.
long Fvstot(char *cstruct, char *trecord, long treclen, char *viewname) long Fvttos(char *cstruct, char *trecord, char *viewname) int Fcodeset(char *translation_table)
The Fvstot function transfers data from a C structure to a target record type. The Fvttos function transfers data from a target record to a C structure. trecord is a pointer to the target record. cstruct is a pointer to a C structure. viewname is a pointer to the name of a compiled view description. The VIEWDIR and VIEWFILES are used to find the directory and file containing the compiled view description.
To convert from an FML buffer to a target record, first call Fvftos to convert the FML buffer to a C structure, and call Fvstot to convert to a target record. To convert from a target record to an FML buffer, first call Fvttos to convert to a C structure and then call Fvstof to convert the structure to an FML buffer.
The default target is IBM/370 COBOL records. The default data conversion is done based on the following table.
| Struct | Record |
|---|---|
| float | COMP-1 |
| double | COMP-2 |
| long | S9(9) COMP |
| short | S9(4) COMP |
| int | S9(9) COMP or S9(4) COMP |
| dec_t(m, n) | S9(2*m-(n+1))V9(n)COMP-3 |
| ASCII char | EBCDIC char |
| ASCII string | EBCDIC string |
| carray | character array |
No filler bytes are provided between fields in the IBM/370 record. The COBOL SYNC clause should not be specified for any data items that are a part of the structure corresponding to the view. An integer field is converted to either a four or two-byte integer depending on the size of integers on the machine on which the conversion is done. A string field in the view must be terminated with a null when converting to/from the IBM/370 format. The data in a carray field is passed unchanged; no data translation is performed.
Packed decimals exist in the IBM/370 environment as two decimal digits packed into one byte with the low-order half byte used to store the sign. The length of a packed decimal may be 1 to 16 bytes with storage available for 1 to 31 digits and a sign. Packed decimals are supported in C structures using the dec_t field type. The dec_t field has a defined size consisting of two numbers separated by a comma. The number to the left of the comma is the total number of bytes that the decimal occupies. The number to the right is the number of digits to the right of the decimal point. The formula for conversion is:
dec_t(m, n) <=> S9(2*m-(n+1))V9(n)COMP-3
Decimal values may be converted to and from other data types (e.g., int, long, string, double, and float) using the functions described in decimal(3c).
See the Fvstof(3) manual page for the default character conversion of ASCII to/from EBCDIC.
An alternate character translation table can be used at run-time by calling Fcodeset. The translation_table must point to 512 bytes of binary data. The first 256 bytes of data are interpreted as the ASCII to EBCDIC translation table. The second 256 bytes of data are interpreted as the EBCDIC to ASCII table. Any data after the 512th byte is ignored. If the pointer is NULL, the default translation is used.