System APIs use a language-neutral convention when defining the type and size of data that is to be used as a parameter or a subfield of a data structure. A typical API parameter list would be the Retrieve Member Description QUSRMBRD API shown below and documented here.
Retrieve Member Description (QUSRMBRD) API
Required Parameter Group:
1 |
Receiver variable |
Output |
Char(*) |
2 |
Length of receiver variable |
Input |
Binary(4) |
3 |
Format name |
Input |
Char(8) |
4 |
Qualified database file name |
Input |
Char(20) |
5 |
Database member name |
Input |
Char(10) |
6 |
Override processing |
Input |
Char(1) |
Optional Parameter Group 1:
7 |
Error code |
I/O |
Char(*) |
Optional Parameter Group 2:
8 |
Find member processing |
Input |
Char(1) |
The API documentation, in the fourth column above, is defining the attributes associated with various parameters that are to be passed to the API. Most system APIs utilize only two types of data (character and binary) and define the size, or length, of the data in terms of bytes. So in the case of character data, the third parameter of QUSRMBRD, Format name, is defined as a Char (or character) field with a fixed length of eight bytes. Likewise, the fourth parameter, Qualified database file name, is a Char field with a fixed length of 20 bytes and the sixth parameter, Override processing, a Char field of one byte. You'll notice that two of the parameters, Receiver variable and Error code, do not specify a fixed length. Rather, they use a length of '*'. The '*' is a special value indicating that the parameter is of variable length and that you, the caller of the API, control (within certain boundaries) the actual size of the parameter being passed. Quite often parameters defined as Char(*) are actually data structures and within the data structure will be fixed-size subfields. You will also find in the API documentation that some of the fixed-length Char parameters are also data structures. Qualified database file name, for instance, is actually a data structure with two subfields: a Char(10) file name and a Char(10) library name.
The second parameter, Length of receiver variable, is defined as a Binary field with a fixed length of four bytes. This definition for a binary, or integer, field can be a source of confusion to RPG developers and is a common pitfall when first using APIs. The pitfall is that the RPG developer may code the parameter/field using a number of digits specification as shown below:
DName+++++++++++ETDsFrom+++To/L+++IDc.Keywords++++++++
dLenRcvVar s 4b 0
This definition "looks" right but, unfortunately, is wrong. What is being defined above is a binary field capable of holding a numeric value ranging from 1 to 4 digits in size. As a binary format can represent four digits in two bytes, RPG will only allocate two bytes for the field LenRcvVar (Length of receiver variable). As the API expects four bytes to be passed, this is not going to work very well, and rather unpredictable errors can occur at run time. The correct definition of a Binary(4) can be done using any one of the below statements:
DName+++++++++++ETDsFrom+++To/L+++IDc.Keywords++++++++
* Standalone fields
dLenRcvVar s 9b 0
dLenRcvVar s 10i 0 (A)
* Data structure subfields
d LenRcvVar 1 4b 0
d LenRcvVar 1 4i 0 (B)
d LenRcvVar 9b 0
d LenRcvVar 10i 0 (C)
I'll point out that the use of the 'i' data type (examples A, B, and C above) is the preferred method. And if it helps any, C developers when first using system APIs also run into a common pitfall: They tend to interpret Char(*) as a pointer to a character string, which also "looks" right to them, but isn't.
As this binary definition confusion is a reasonably common problem for RPG developers, it has been suggested in the past that IBM change the documentation of binary fields from Binary(4) to something more RPGish. While IBM may make such a change in the future, my hope is that the documentation (at least in this regard) remains the same. One reason for this hope is that RPG continues to evolve. If back in V1R3, when system APIs were first introduced, IBM had used a convention such as Binary(9) to represent a 4-byte binary field, then V3R2/V3R6 and ILE RPG's introduction of the 4-byte integer data type might have introduced an unnecessary dilemma: namely, should IBM change the API documentation to Integer(10) to reflect the new RPG definition style? If so, wouldn't the pitfall discussed above simply be reintroduced for RPG/400 and ILE users of the binary data type? Binary(10) or Integer(10) would also certainly be more accurate as developers familiar with binary data would know that a 4-byte binary field can indeed represent 10 digits of information. It's only for historical reasons that RPG (and some other parts of the system) constrains 4-byte binary definitions to nine digits.
Related to this, having a language-neutral specification for API documentation also avoids the false perception of the System i being a RPG machine: a marketing perception any fan of the System i would want to avoid in order to enhance acceptance of the system in the marketplace. Other languages that are generally available for application development on i5/OS define a 4-byte integer field in various ways, none of which by the way are Binary(4)! In COBOL, you might use 'PIC S9(9) BINARY', in C 'int', in CL (or at least in current releases of CL) 'Type(*Int) Len(4)', and so on. Using a neutral definition such as Binary(4) puts all development languages on a somewhat level field; everyone has to learn what it means.
To digress a little, this type of language neutrality reminds me of the "acronym" UTC for Coordinated Universal Time. UTC is not a true acronym; it's a compromise between the English Coordinated Universal Time (CUT) and French Temps Universel Coordonné (TUC). And so Binary(4) is a compromise between the various programming languages available today and in the future.
The good thing is that once learned, it's very easy to remember.
An API data structure is defined using a similar style. Below is the API documentation for the MBRD0100 format of the QUSRMBRD API.
MBRD0100 Format
Offset |
Type |
Field |
|
Dec |
Hex |
||
0 |
0 |
BINARY(4) |
Bytes returned |
4 |
4 |
BINARY(4) |
Bytes available |
8 |
8 |
CHAR(10) |
Database file name |
18 |
12 |
CHAR(10) |
Database file library name |
28 |
1C |
CHAR(10) |
Member name |
38 |
26 |
CHAR(10) |
File attribute |
48 |
30 |
CHAR(10) |
Source type |
58 |
3A |
CHAR(13) |
Creation date and time |
71 |
47 |
CHAR(13) |
Last source change or table refresh date and time |
84 |
54 |
CHAR(50) |
Member text description |
134 |
86 |
CHAR(1) |
Source file |
As you can hopefully see, the knowledge acquired about parameter data types is easily transferred to data structure data types. There are a few additional data types that you may encounter over time. These include pointers (a data type of PTR and implicitly defined as 16-bytes); Binary(2) fields, which you should successfully interpret as 2-byte binary fields after reading this column; and Packed(x,y) for packed decimal fields with x digits, y digits of which are to the right of the decimal point. These additional types do not however come up very often.
With this introduction to API data types, you should be well on your way toward successfully understanding how to utilize i5/OS system API documentation.
Meanwhile, if you have other API questions, send them to Bruce at
LATEST COMMENTS
MC Press Online