GdaSqlParser

GdaSqlParser — SQL parser

Stability Level

Stable, unless otherwise indicated

Properties

gint column-error Read
gboolean debug Write
gint line-error Read
gint mode Read / Write
gint tokenizer-flavour Read / Write

Types and Values

Object Hierarchy

    GObject
    ╰── GdaSqlParser

Implemented Interfaces

GdaSqlParser implements GdaLockable.

Includes

#include <sql-parser/gda-sql-parser.h>

Description

The GdaSqlParser is an object dedicated to creating GdaStatement and GdaBatch objects from SQL strings. The actual contents of the parsed statements is represented as GdaSqlStatement structures (which can be obtained from any GdaStatement through the "structure" property).

GdaSqlParser parsers can be created by calling gda_server_provider_create_parser() for a provider adapted SQL parser, or using gda_sql_parser_new() for a general purpose SQL parser.

The GdaSqlParser can either work in "parse" mode where it will try to parse the SQL string, or in "delimiter" mode where it will only attempt at delimiting SQL statements in a string which may contain several SQL statements (usually separated by a semi column). If operating in "parser" mode, and the parser can't correctly parse the string, then it will switch to the "delimiter" mode for the next statement in the string to parse (and create a GDA_SQL_STATEMENT_UNKNOWN statement).

The GdaSqlParser object parses and analyzes SQL statements and reports the following statement types:

  • SELECT (and COMPOUND select), INSERT, UPDATE and DELETE SQL statements should be completely parsed.

  • Transaction related statements (corresponding to the BEGIN, COMMIT, ROLLBACK, SAVEPOINT, ROLLBACK SAVEPOINT and DELETE SAVEPOINT) are parsed and a minimalist structure is created to extract some information (that structure is not enough per-se to re-create the complete SQL statement).

  • Any other type of SQL statement (CREATE TABLE, ...) creates a GdaStatement of type GDA_SQL_STATEMENT_UNKNOWN, and it only able to locate place holders (variables) and end of statement marks.

NOTE: Any SQL of a type which should be parsed which but which creates a GdaStatement of type GDA_SQL_STATEMENT_UNKNOWN (check with gda_statement_get_statement_type()) should be reported as a bug.

The GdaSqlParser object recognizes place holders (variables), which can later be queried and valued using gda_statement_get_parameters(). The following syntax are recognized (other syntaxes might be recognized for specific database providers if the GdaSqlParser is created using gda_server_provider_create_parser() but for portability reasons it's better to avoid them):

  • #NAME[::TYPE[::NULL]]]

    : for a variable named NAME with the optional type TYPE (which can be a GType name or a custom database type name), and with the optional "::NULL" to instruct that the variable can be NULL.

  • ## /* name:NAME [type:TYPE] [nullok:[TRUE|FALSE]] [descr:DESCR] */

    for a variable named NAME with the optional type TYPE (which can be a GType name or a custom database type name), with the optional "nullok" attribute and an optional description DESCR. Note that the NAME, TYPE and DESCR literals here must be quoted (simple or double quotes) if they include non alphanumeric characters, and that there must always be at least a space between the ## and the opening and closing comments (C style).

Note that the type string must be a type recognized by the

gda_g_type_from_string() function (all valid GType names

plus a few synonyms). Examples of correct place holders definitions are:

## /* name:"+0" type:gchararray */
## /* name:'-5' type:string */
## /*name:myvar type:gint descr:ToBeDefined nullok:FALSE*/
## /*name:myvar type:int descr:"A long description"*/
##+0::gchararray
##-5::timestamp

Also note that variables should not be used when an SQL identifier is expected. For example the following examples should be avoided because they may not work properly (depending on the database being used):

SELECT * FROM #tablename::string;
DELETE FROM mytable WHERE #tcol::string = 5;
ALTER GROUP mygroup ADD USER #name::gchararray;

The GdaSqlParser object internally uses a LEMON generated parser (the same as the one used by SQLite).

The GdaSqlParser object implements its own locking mechanism so it is thread-safe.

Functions

gda_sql_parser_new ()

GdaSqlParser *
gda_sql_parser_new (void);

Creates a new GdaSqlParser object

Returns

the new object


gda_sql_parser_parse_string ()

GdaStatement *
gda_sql_parser_parse_string (GdaSqlParser *parser,
                             const gchar *sql,
                             const gchar **remain,
                             GError **error);

Parses sql and creates a GdaStatement statement from the first SQL statement contained in sql : if sql contains more than one statement, then the remaining part of the string is not parsed at all, and remain (if not NULL) will point at the first non parsed character.

To include variables in the sql string, see the

GdaSqlParser's object description.

Parameters

parser

a GdaSqlParser object

 

sql

the SQL string to parse

 

remain

location to store a pointer to remaining part of sql in case sql has multiple statement, or NULL.

[out][allow-none]

error

location to store error, or NULL

 

Returns

a new GdaStatement object, or NULL if an error occurred.

[transfer full][allow-none]


gda_sql_parser_parse_string_as_batch ()

GdaBatch *
gda_sql_parser_parse_string_as_batch (GdaSqlParser *parser,
                                      const gchar *sql,
                                      const gchar **remain,
                                      GError **error);

Parse sql and creates a GdaBatch object which contains all the GdaStatement objects created while parsing (one object per SQL statement). Empty statements (composed of spaces only) do not appear in the resulting object.

sql is parsed and GdaStatement objects are created as long as no error is found in sql . If an error is found at some point, then the parsing stops and remain may contain a non NULL pointer, error may be set, and NULL is returned.

if sql is NULL, then the returned GdaBatch object will contain no statement.

To include variables in the sql string, see the

GdaSqlParser's object description.

Parameters

parser

a GdaSqlParser object

 

sql

the SQL string to parse

 

remain

location to store a pointer to remaining part of sql in case an error occurred while parsing sql , or NULL.

[out][allow-none]

error

location to store error, or NULL

 

Returns

a new GdaBatch object, or NULL if an error occurred.

[transfer full][allow-none]


gda_sql_parser_parse_file_as_batch ()

GdaBatch *
gda_sql_parser_parse_file_as_batch (GdaSqlParser *parser,
                                    const gchar *filename,
                                    GError **error);

Parse filename 's contents and creates a GdaBatch object which contains all the GdaStatement objects created while parsing (one object per SQL statement).

filename 's contents are parsed and GdaStatement objects are created as long as no error is found. If an error is found at some point, then the parsing stops, error may be set and NULL is returned

if sql is NULL, then the returned GdaBatch object will contain no statement.

Parameters

parser

a GdaSqlParser object

 

filename

name of the file to parse

 

error

location to store error, or NULL

 

Returns

a new GdaBatch object, or NULL if an error occurred.

[transfer full][allow-none]


gda_sql_identifier_quote ()

gchar *
gda_sql_identifier_quote (const gchar *id,
                          GdaConnection *cnc,
                          GdaServerProvider *prov,
                          gboolean meta_store_convention,
                          gboolean force_quotes);

Use this function for any SQL identifier to make sure that:

  • it is correctly formatted to be used with cnc (if cnc is NULL, then some default SQL quoting rules will be applied, similar to PostgreSQL's way) if for_meta_store is FALSE;

  • it is correctly formatted to be used with the GdaMetaStore's object associated to cnc is for_meta_store is TRUE.

The force_quotes allow some control of how to interpret id : if FALSE, then id will be left unchanged most of the time (except for example if it's a reserved keyword), otherwise if force_quotes is TRUE, then the returned string will most probably have quotes around it to request that the database keep the case sensitiveness (but again, this may vary depending on the database being accessed through cnc ).

For example, the following table gives the result of this function depending on the arguments when cnc is NULL (and prov is also NULL):

Table 3. 

id for_meta_store=FALSE, force_quotes=FALSE for_meta_store=TRUE, force_quotes=FALSE for_meta_store=FALSE, force_quotes=TRUE for_meta_store=TRUE, force_quotes=TRUE remark
"double word" "double word" "double word" "double word" "double word" non allowed character in SQL identifier
"CapitalTest" "CapitalTest" "CapitalTest" "CapitalTest" "CapitalTest" Mixed case SQL identifier, already quoted
CapitalTest CapitalTest capitaltest "CapitalTest" "CapitalTest" Mixed case SQL identifier, non quoted
"mytable" "mytable" mytable "mytable" mytable All lowser case, quoted
mytable mytable mytable "mytable" mytable All lowser case
MYTABLE MYTABLE mytable "MYTABLE" "MYTABLE" All upper case
"MYTABLE" "MYTABLE" "MYTABLE" "MYTABLE" "MYTABLE" All upper case, quoted
desc "desc" "desc" "desc" "desc" SQL reserved keyword
5ive "5ive" "5ive" "5ive" "5ive" SQL identifier starting with a digit


Here are a few examples of when and how to use this function:

  • When creating a table, the user has entered the table name, this function can be used to create a valid SQL identifier from the user provided table name:

    gchar *user_sqlid=...
    gchar *valid_sqlid = gda_sql_identifier_quote (user_sqlid, cnc, NULL, FALSE, FALSE);
    gchar *sql = g_strdup_printf ("CREATE TABLE %s ...", valid_sqlid);
    g_free (valid_sqlid);
          

    Note that this is an illustration and creating a table should be sone using a GdaServerOperation object.

  • When updating the meta data associated to a table which has been created with the code above:

    GValue table_name_value = { 0 };
    gchar* column_names[] = { (gchar*)"table_name" };
    GValue* column_values[] = { &table_name_value };
    GdaMetaContext mcontext = { (gchar*)"_tables", 1, column_names, column_values };
    g_value_init (&table_name_value, G_TYPE_STRING);
    g_value_take_string (&table_name_value, gda_sql_identifier_quote (user_sqlid, cnc, NULL, TRUE, FALSE);
    gda_connection_update_meta_store (cnc, &mcontext, NULL);
    g_value_reset (&table_name_value);
          

  • When using a GdaMetaStruct object to fetch information about a table (which has been created with the code above):

    GValue table_name_value = { 0 };
    g_value_init (&table_name_value, G_TYPE_STRING);
    g_value_take_string (&table_name_value, gda_sql_identifier_quote (user_sqlid, cnc, NULL, TRUE, FALSE);
    GdaMetaDbObject *dbo;
    dbo = gda_meta_struct_complement (mstruct, GDA_META_DB_TABLE, NULL, NULL, &table_name_value, NULL);
    g_value_reset (&table_name_value);
          

Note that id must not be a composed SQL identifier (such as "mytable.mycolumn" which should be treated as the "mytable" and "mycolumn" SQL identifiers). If unsure, use gda_sql_identifier_split().

Also note that if cnc is NULL, then it's possible to pass an non NULL prov to have a result specific to prov .

For more information, see the SQL identifiers and abstraction and

SQL identifiers in meta data sections.

Parameters

id

an SQL identifier

 

cnc

a GdaConnection object, or NULL.

[allow-none]

prov

a GdaServerProvider object, or NULL for_meta_store set to TRUE if the returned string will be used in a GdaMetaStore.

[allow-none]

force_quotes

set to TRUE to force the returned string to be quoted

 

Returns

the representation of id ready to be used in SQL statement, as a new string, or NULL if id is in a wrong format

Since 4.0.3


gda_sql_identifier_split ()

gchar **
gda_sql_identifier_split (const gchar *id);

Splits id into an array of it sub parts. id 's format has to be "<part>[.<part>[...]]" where each part is either a text surrounded by double quotes which can contain upper and lower cases or an SQL identifier in lower case.

For example the "test.\"ATable\"" string will result in the array: {"test", "\"ATable\"", NULL}

Parameters

id

an SQL identifier

 

Returns

a new NULL-terminated array of strings, or NULL (use g_strfreev() to free the returned array).

[transfer full][array zero-terminated=1][allow-none]

Types and Values

GdaSqlParser

typedef struct _GdaSqlParser GdaSqlParser;

enum GdaSqlParserMode

Members

GDA_SQL_PARSER_MODE_PARSE

   

GDA_SQL_PARSER_MODE_DELIMIT

   

Property Details

The “column-error” property

  “column-error”             gint

Flags: Read

Allowed values: >= 0

Default value: 0


The “debug” property

  “debug”                    gboolean

Flags: Write

Default value: FALSE


The “line-error” property

  “line-error”               gint

Flags: Read

Allowed values: >= 0

Default value: 0


The “mode” property

  “mode”                     gint

Flags: Read / Write

Allowed values: [0,1]

Default value: 0


The “tokenizer-flavour” property

  “tokenizer-flavour”        gint

Flags: Read / Write

Allowed values: [0,4]

Default value: 0