Chapter 40 – SQL/CLI Binding Style


You are reading a digital copy of SQL-99 Complete, Really, a book that documents the SQL-99 standard.

The book and the standard does not reflect the features of CrateDB, we are just publishing it as a service for the community and for reference purposes.

On the one hand, CrateDB does not implement the SQL-99 standard thoroughly, on the other hand, it extends the standard for implementing cluster features and others.

For more information specific to CrateDB, check out the CrateDB Reference documentation.

SQL DBMSs communicate with SQL applications through a common programming language interface that is invoked through one of the SQL Standard-defined binding styles, or interface options. There are three main approaches to writing complete programs with SQL:

  1. With embedded SQL, you can put SQL statements directly into host programs. We described this binding style in the last chapter.

  2. With the Module binding style, you can dispense with host programs and write entire Modules in SQL. You’ll still have to call these Modules from one of the standard host languages though. We’ll describe this binding style last.

  3. With SQL/CLI, you can call a SQL DBMS’s library from a host program. The SQL statements you want to execute are parameters of the call. Because SQL’s modus operandi won’t quite match the host language’s, helper functions are required for the interface definition. This binding style is the subject of the next several chapters.

The SQL/CLI, or Call-level Interface binding style, is defined in part three of the SQL Standard. In this chapter, we’ll give you an introduction to this SQL binding style.

Embedded SQL used to be the most important way of putting SQL statements into application programs. Nowadays, the CLI is the most important. Not because it’s simpler – far from it! The CLI’s real advantages are:

  • There’s no precompiler (this makes debugging easier).

  • DLLs can be substituted (bringing the usual advantages of modularity).

  • ODBC – the best-known SQL CLI – is very popular.

Onward, then, to the CLI essentials. CLI stands for Call Level Interface. You have probably heard of an API (Application Programming Interface), which is a non-SQL name for the same sort of thing. The CLI defines a set of public functions which can be called from a host language. Each function has a name, a parameter list and a required algorithm (what the DBMS must do when you call it using this function).

Table of Contents


Here’s a short C program that uses the CLI (the prototypes and short code examples that follow throughout our description of the CLI are all in C):

#include "sqlcli.h"                                             /* [Note 1] */
SQLHENV     henv;                                               /* [Note 2] */
SQLHDBC     hdbc;
SQLHSTMT    hstmt;
void main ()
  SQLAllocHandle(SQL_HANDLE_ENV,NULL,&henv);                    /* [Note 3] */
  SQLExecDirect(hstmt,"CREATE TABLE World(hello INT);",SQL_NTS);/* [Note 4] */
  SQLFreeHandle(SQL_HANDLE_STMT,hstmt);                         /* [Note 5] */
  SQLFreeHandle(SQL_HANDLE_ENV,henv); }

This sample program has all the necessary CLI ingredients. It will take several pages to say how they all work, but the main points are noted in the code as follows:

[Note 1] Include the SQL/CLI header file. The standard name is "sqlcli.h" but you might see this statement – #include "sql.h" – instead. The header file has the prototypes of all the CLI functions and #defines for all the constants. The function names begin with SQL and the constant names begin with SQL_ – that’s the convention in C programs. You can program standard SQL without following this convention, but it seems reasonable to follow it.

[Note 2] Variable declarations. The convention this time is: h stands for handle – so a henv is a handle of an env, a hdbc is a handle of a dbc and a hstmt is a handle of a stmt. And now, since those abbreviated words will appear many times in the next 200 pages, we’d advise you to memorize them:

  • env is an “allocated thing [resource] used for an environment”.

  • dbc is an “allocated thing [resource] used for a database connection”.

  • stmt is an “allocated thing [resource] used for a statement”.

For the record, the Standard’s usual names for env, dbc and stmt are “allocated SQL-environment”, “allocated SQL-connection” and “allocated SQL-statement”, respectively. We will always use the abbreviations because they are what appear in sqlcli.h, and in all programs. Besides, we believe it’s crazy to call a stmt a statement – a stmt is a RESOURCE; a statement is a STRING. Let’s not confuse the peanut with the shell. As it happens, the sqlcli.h header file has “typedef”s for SQLHENV and the other handles: they’re all 32-bit integers. Other important typedefs are SQLINTEGER (“long int”), SQLSMALLINT (“short int”), SQLCHAR (“char”) and SQLRETURN (“short int”, used for return codes only). Using, say, “SQLINTEGER x” in a program rather than “long int x” helps make it clear that “x” will see use in an SQL function.

[Note 3] Setup functions. The bare minimum procedure involves making an env, making a dbc, connecting using the dbc and making a stmt. Notice the &s in the program code: these are “output” parameters (“output” from the DBMS point of view). They’re passed by address because the DBMS fills in the values for them.

[Note 4] Meat. The only actual SQL-language statement in our program is a string argument for a CLI function call. This is the actual database work.

[Note 5] Tear-down functions. In a reverse of the setup procedure, we call functions for destroying a stmt, disconnecting using the dbc, destroying the dbc and destroying the env. Notice that every function call used a handle – that’s normal. Doubtless the DBMS is merely malloc’ing a bit of memory for each of our “allocated things”, but we can’t access the fields directly. Instead, we use a henv or a hdbc or a hstmt.

SQLCHAR, SQLINTEGER and Other Typedefs

Here are the type definitions in the sqlcli.h header file. We use these names for declarations of C variables in all our examples.

typedef unsigned char  SQLCHAR;      /* 8-bit-octet strings */
typedef long int       SQLINTEGER;   /* 32-bit, signed */
typedef short int      SQLSMALLINT;  /* 16-bit, signed */
typedef float          SQLREAL;      /* see heading: IEEE */
typedef double         SQLDOUBLE;    /* see heading: IEEE */
typedef void*          SQLPOINTER;   /* pointer, untyped */
typedef long int       SQLHENV;      /* 32-bit env handle */
typedef long int       SQLHDBC;      /* 32-bit dbc handle */
typedef long int       SQLHSTMT;     /* 32-bit hstmt handle */
typedef long int       SQLHDESC;     /* 32-bit desc handle */


In the ODBC 3.0 header file, SQLHENV and SQLHDBC and SQLHSTMT and SQLHDESC are all typedef void* instead of typedef long int. In older versions of the ODBC header file, the names were HENV, HDBC, HSTMT, etc.


All CLI functions return a 16-bit value. We refer to this value as the SQLRETURN value because sqlcli.h contains this line:


In standard SQL there are only six possible SQLRETURN values:


“invalid handle”.






“success with info”.


“need data”.


“no data”.

Programs should check these values, but for space reasons we leave SQLRETURN out of many of our examples.

Handle Relationships

A handle is a 32-bit integer which can be used as a unique identifier of a “resource”. The following chart shows all the resources which have handles. The relationships between resources are either optional (“zero-to-many” – shown with double arrows) or mandatory (either “one-to-one” or “zero-to-one” – shown with single arrows).

dbc --------------->
||                 ||
||                 vv
||                 desc
|       |       |       |
v       v       v       v
ARD     APD     IRD     IPD
(desc)  (desc)  (desc)  (desc)

How to Run Example Programs

This book contains several example programs that we use to illustrate the SQL/CLI. Each example program is also included on the CD that came with this book, numbered in order of appearance (for example the “example C program” shown earlier is example1.c). To try out any of the example programs, follow these steps.

If you use Microsoft Windows 3.x:

  • Copy these files from the CD to your working directory: OCELOT16.DLL, OCELOT16.LIB, SQLCLI.H and the C program file.

  • Using a standard-C package, compile and link. For eample, this is a sequence for Borland C++ version 3.1 (with the MS-DOS command line):

bcc /I\borlandc\include /L\borlandc\lib /W /M /N /ml example1.c ocelot16.lib
  • Run the program from the MS-DOS command line.

If you use Microsoft Windows 95:

  • Copy these files from the diskette to your working directory: OCELOT32.DLL, OCELOT32.LIB, SQLCLI.H and the C program file.

  • Using a standard-C package, compile and link. For example, this is a sequence for Symantec C++ version 7.0 (with the MS-DOS command line):

sc /I\sc\include /WA /g /mn /s /L/M example1.c kernel32.lib ocelot32.lib
  • Run the program from the MS-DOS command line.


Add your own screen-display statements so you’ll see the values of program variables when you run the program. If there’s a CLI function that you want to get experience with, put it into your own program, link with the library supplied with this book, and run. As well as being good practice, this will help you learn what is NOT standard SQL, since we deliberately kept out “implementation-defined” features when we wrote this DBMS.

“Standard SQL CLI” Equals “Core ODBC API”

History: Microsoft published the ODBC 1.0 specification in 1992. It became popular. Soon there were supporting interfaces for all the major DBMSs that worked under Microsoft Windows. ODBC became a de facto standard specification for calling SQL from host programs without a precompiler. The ISO Standard CLI came out in 1995. The influence of ODBC on the standard CLI is apparent in almost every routine. Indeed, it sometimes is hard to understand the official Standard document without studying Microsoft’s ODBC manual too. On its part, Microsoft made major changes in ODBC 3.0 (1998). They added several new functions and deprecated several old ones, in order to get closer to the SQL Standard. Microsoft’s ODBC 3.0 manual makes these claims:

“ODBC aligns with the following specifications and standards that deal with the Call-Level Interface (CLI). (ODBC’s features are a superset of each of these standards.)

  • The X/Open CAE Specification “Data Management: SQL Call-Level Interface (CLI)

  • ISO/IEC 9075-3:1995 (E) Call-Level Interface (SQL/CLI)

As a result of this alignment, the following are true:

  • An application written to the X/Open and ISO CLI specifications will work with an ODBC 3.0 driver or a standards-compliant driver when it is compiled with the ODBC 3.0 header files and linked with ODBC 3.0 libraries, and when it gains access to the driver through the ODBC 3.0 Driver Manager.

  • A driver written to the X/Open and ISO CLI specifications will work with an ODBC 3.0 application or a standards-compliant application when it is compiled with the ODBC 3.0 header files and linked with ODBC 3.0 libraries, and when the application gains access to the driver through the ODBC 3.0 Driver Manager.

The Core interface conformance level encompasses all the features in the ISO CLI and all the non-optional features in the X/Open CLI. Optional features of the X/Open CLI appear in higher interface conformance levels. Because all ODBC 3.0 drivers are required to support the features in the Core interface conformance level, the following are true:

  • An ODBC 3.0 driver will support all the features used by a standards-compliant application.

  • An ODBC 3.0 application using only the features in ISO CLI and the non-optional features of the X/Open CLI will work with any standards-compliant driver.”

Whoa! The SQL Standard CLI isn’t quite as close to Core ODBC as Microsoft is suggesting. There are a few incompatibilities in names, there are two differences in prototype definitions (of SQLColumnAttr and SQLGetInfo) and ODBC’s default behaviour for “commits” is sui generis. But we’ll note those bugs as we step on them. Once you know what they are, you’ll be able to adjust.

Since this is a book on the SQL Standard, our chapters on the CLI won’t show you the various extra features that appear in ODBC but not in the standard CLI. And we note only once – now – that any feature which is marked “SQL3” is a feature of the standard CLI but not necessarily of ODBC.

How Each CLI Function Will Be Described

In these chapters, we will describe each CLI function separately. We’ve put them in order according to this rule: if we can’t describe A without knowing about B, then B should come before A. This means that some tedious minor functions precede some important ones, but the advantage is that you can read from front to back. Better to slog, rather than jump around.

The structure of each CLI-function description is semi-rigid. You can expect to find these things, in this order:

  1. Function prototype. This is the prototype as it appears in the header file, sqlcli.h. It gives a quick idea of what the function’s name is and what <data type>s the parameters have. There is a comment beside each parameter indicating whether it’s an “input” parameter (its value goes from the application to the DBMS) or an “output” parameter (its value goes from the DBMS to the application).

  2. Job. A one-sentence description of what the function is for. This may be followed by an indication of whether the function is essential or not. Several functions are obsolete or redundant, so you can skip the details on your first pass through these chapters.

  3. Algorithm. Exactly what does the DBMS do with this function? This section is in a sort of pseudocode. We’re trying to make the instructions readable, but we won’t spare you from mind-numbing (but necessary) details.

  4. Notes. Whatever strikes us as worthy of remark about a function.

  5. Example. A code snippet written in C. Usually the example will be an incomplete program, with … to indicate that we haven’t repeated stuff which doesn’t advance the story. Most commonly, we leave out the “setup” and “tear- down” phases illustrated in the example program shown at the beginning of this chapter.

  6. ODBC. Anything that relates specifically to ODBC, but not to the standard CLI. If a standard CLI procedure doesn’t exactly match the ODBC spec, it gets noted here. If ODBC has a truly significant feature that is not in the standard CLI, we mention it curtly.

Having said that, let’s get on with the descriptions. Here’s a quick list of the 62 standard CLI functions:

NAME                 CATEGORY     FUNCTION
SQLAllocConnect      dbc          Obsolescent: Make a dbc
SQLAllocEnv          env          Obsolescent: Make an env
SQLAllocHandle       dbc,env,stmt Essential: Make env, dbc, stmt or desc
SQLAllocStmt         stmt         Obsolescent: Make a stmt
SQLBindCol           desc         Useful: Associate Column descriptor to stmt
SQLBindParameter     desc         Useful: Associate parameter to stmt
SQLCancel            D parameters Minor: Stop unfinished statement execution
SQLCloseCursor       Cursors      Essential: Close cursor
SQLColAttribute      desc         Useful: Get info re result set structure
SQLColumnPrivileges  Catalog      Junk: Get metadata re Column privileges
SQLColumns           Catalog      Junk: Get metadata re Columns
SQLConnect           dbc          Essential: Connect to SQL-server/database
SQLCopyDesc          desc         Minor: Make copy of descriptor
SQLDataSources       general      Minor: List available servers
SQLDescribeCol       desc         Useful: Get info re result set structure
SQLDisconnect        dbc          Essential: Disconnect SQL-server/database
SQLEndTran           Statements   Essential: Commit or Rollback
SQLError             Diagnostics  Obsolete: Get diagnostics information
SQLExecDirect        Statements   Useful: Prepare + Execute a statement
SQLExecute           Statements   Essential: Execute a prepared statement
SQLFetch             Cursors      Useful: Bring in next result-set row
SQLFetchScroll       Cursors      Essential: Bring in any result-set row
SQLForeignKeys       Catalog      Junk: Get metadata re foreign keys
SQLFreeConnect       dbc          Obsolescent: Destroy dbc
SQLFreeEnv           env          Obsolescent: Destroy env
SQLFreeHandle        Handles      Essential: Destroy env, dbc, stmt or desc
SQLFreeStmt          stmt         Obsolescent: Destroy stmt and other things
SQLGetConnectAttr    dbc          Useless: Get attribute of dbc
SQLGetCursorName     Cursors      Minor: Get Cursor name
SQLGetData           desc         Useful: Get value from result set
SQLGetDescField      desc         Essential: Get 1 desc field
SQLGetDescRec        desc         Useful: Get 7 desc fields
SQLGetDiagField      Diagnostics  Essential: Get diagnostics information
SQLGetDiagRec        Diagnostics  Useful:  Get diagnostics information
SQLGetEnvAttr        env          Minor: Get attribute of env
SQLGetFunctions      general      Minor: List CLI functions supported
SQLGetInfo           general      Essential: Get attribute of dbc etc.
SQLGetLength         locator      Minor: Length of CLOB/BLOB
SQLGetParamData      desc         Minor: Get data from procedure parameters
SQLGetPosition       locator      Minor: Indicate substring start position
SQLGetStmtAttr       stmt         Essential: Get attribute of stmt
SQLGetSubstring      locator      Minor: Get portion of CLOB/BLOB
SQLGetTypeInfo       Catalog      Junk: Get metadata re data types
SQLMoreResults       Cursors      Minor: See if more result sets
SQLNumResultCols     desc         Useful: Get selected-Column count
SQLParamData         D parameters Minor: For passing data piecemeal
SQLParameters        Catalog      Junk: Get metadata re parameters
SQLPrepare           Statements   Essential: Prepare a statement
SQLPrimaryKeys       Catalog      Junk: Get metadata re primary keys
SQLPutData           D parameters Minor: For passing data piecemeal
SQLRoutinePrivileges Catalog      Junk: Get metadata re Privileges
SQLRoutines          Catalog      Junk: Get metadata re routines
SQLRowCount          Diagnostics  Useful:  Get affected-rows count
SQLSetConnectAttr    dbc          Useless: Change dbc attribute
SQLSetCursorName     Cursors      Essential: Change Cursor name
SQLSetDescField      desc         Essential:  Change 1 desc field
SQLSetDescRec        desc         Useful: Change 7 desc fields
SQLSetEnvAttr        env          Useless: Change env attribute
SQLSetStmtAttr       stmt         Useful: Change stmt attribute
SQLSpecialColumns    Catalog      Junk: Get metadata re search Columns
SQLTablePrivileges   Catalog      Junk: Get metadata re Table Privileges
SQLTables            Catalog      Junk: Get metadata re Tables

And that concludes our brief introduction to the CLI. Before we begin describing each of the functions in this list in detail, though, we’ll show you a sample function description, using a common CLI function subroutine – CharacterStringRetrieval.


Function Prototype:

void CharacterStringRetrieval (
  SQLCHAR *Target,                  /* CHAR* output */
  SQLCHAR *Source,                  /* CHAR* input */
  SQLINTEGER MaxTargetOctetLength,  /* 32-bit output */
  SQLINTEGER *SourceOctetLength     /* 32-bit output */

Job: The common subroutine for Character String Retrieval appears in 24 CLI functions. You don’t directly call it, but you see the effects. We thought this would be a good way to introduce the format we use for CLI function descriptions.


If (MaxTargetOctetLength <= 0)
  return error: HY090 CLI-specific condition-invalid string length or buffer length
If (SourceOctetLength is not a null pointer)
  /* We set *SourceOctetLength = the size of the source string, even if we don't actually copy the entire source to target. */
  Set *ReturnedOctetLength = strlen (Value)
If (we don't use null-termination in this environment)
  /* In COBOL or FORTRAN you wouldn't expect null-termination, since strings don't end with a '\0' in those languages. But we never take this path. */
  /* In C, and also in Pascal, strings end with '\0'. We'll assume null-termination then. But now you're aware that it's a characteristic of the host language we're using, not a requirement of standard SQL. */
  If (we're using WideChar i.e. 16 bits per character)
    sizeof(null-terminator) = 2
  Else sizeof(null-terminator) = 1
  Set # of octets to copy = MaxTargetOctetLength-sizeof(null-terminator)
  If (# of octets to copy < 0)
    /* Put out '\0' if possible, but there's no room for copying */
    If (# of octets to copy > strlen(Source))
      Set # of octets to copy = strlen(source)
        Set # of octets to copy=strlen(Source) - sizeof(null-terminator)
        From: Source
        To: Target
    Append null-terminator to Target.
  If (strlen(Target) < Strlen(Source)
    warning - string data, right truncation


  • The idea is simply to copy the *Source string to the *Target string, but there’s a safeguard against overflowing *Target (that’s why we pass MaxTargetOctetLength) and there’s a place to store the original size (that’s why we pass *ReturnedOctetLength).

  • The *Target string will always be null-terminated, unless TargetOctetLength=0.

  • There are variations of this routine where ...OctetLength is 16-bit.

  • There is a slight variation of this routine for BLOB retrieval: it works the same way, but ignores anything to do with null terminators.

  • CharacterStringRetrieval(target,source,max,&sourcelength) is expressible using the standard-C <string.h> function strxfrm: sourcelength = strxfrm(target,source,max);


/ * SPECIMEN ONLY -- you don't really want to call this function */
#include "sqlcli.h"     /* includes definitions of SQL... stuff */
SQLCHAR target[100];    /* SQLCHAR is 8-bit ISO if Windows98 */
SQLINTEGER returnsize;  /* SQLINTEGER is always 32-bit signed */
/* Now target has: ABC\0 and returnsize = 5. */
/* Now target has: AB\0 and returnsize = 2. */

ODBC: In ODBC 1.0’s documentation there was some contradictory information about the workings of Character String Retrieval, but everything is settled now.