In case of very large compound selects or an INSERT with many
values, qc_sqlite could run out of stack space. To deal with
that, the critical recursion is turned into an iteration.
There are some issues still
- With recursive CTEs qc_mysqlembedded and qc_sqlite agree upon
the real columns, but disagree on the CTE related "virtual"
columns. That's largely irrelevant though.
- qc_sqlite cannot parse "SET STATEMENT var=... FOR stmt",
but it will be classified as QUERY_TYPE_GSYSVAR_WRITE.
This is not directly CTE related.
Alias handling must be made so that in a subselect, aliases created
in an outer select are available, but aliases created in another
subselect are not.
The sqlite3 initialization is done a bit more properly now.
It is also ensured that issues are logged at most once, even
if a statement is parsed twice.
The "parser" of compare cannot handle the following:
prepare stmt1 from "
with t as (select a from t1 where b >= 'c')
select * from t2,t where t2.c=t.a;
";
That is, that a line *not* terminating the full statement (the
line 'select * from t2,t where t2.c=t.a;' above) ends with a ';'.
Thus, it needs to be changed into e.g.
prepare stmt1 from "
with t as (select a from t1 where b >= 'c')
select * from t2,t where t2.c=t.a;";
In the context of subselects and CTEs it is somewhat difficult to
make qc_sqlite and qc_mysqlembedded to agree upon in what context
a particular field is used. As that information is not used anywhere,
a usage discrepancy is for now treated as a warning.
All callbacks called by sqlite now only access the thread specific
QcSqliteInfo and call the corresponding function on that instance.
This is the first step in making qc_sqlite exception safe from the
point of view of sqlite3.
As the diff is confusing, basically the ONLY thing that has been
done is:
BEFORE:
-------
class QcSqliteInfo
{
...
};
static void some_helper(...) { ... }
void mxs_someCallback(...)
{
QC_TRACE();
QcSqliteInfo* info = this_thread.pInfo;
ss_dassert(info);
info->m_status = ...;
some_helper(info, ...);
}
AFTER:
------
class QcSqliteInfo
{
...
void some_helper(...) { ... }
void mxs_someCallback(...)
{
m_status = ...;
some_helper(this, ...);
}
};
void mxs_someCallback(...)
{
QC_TRACE();
QcSqliteInfo* pInfo = this_thread.pInfo;
ss_dassert(pInfo);
pInfo->mxs_someCallback(...);
}
With this change, for a statement like
SELECT t2.a FROM t1 t2;
the affected field is reported as t1.a and not as t2.a, as it
was before.
For a statement like
SELECT t.f FROM d.t;
qc_mysqlembedded will now return "d.t.f" as the affected field,
while qc_sqlite will still return "t.f" as both implementations did
before. In qc_mysqlembedded's case that is a side-effect of the
alias handling. To get qc_sqlite to return the same (which would
be good), the table names would have to be collected in a smarter
way than they are now.
Alias names for tables are now collected, so that the true
table name of a referred to field can be reported. That modification
will be made in a subsequent commit.
If a particular table appears in a statement multiple times,
qc_get_table_names will report it as many times as it appears.
Each name should be reported just once. Same applies for
database names.
The update_names() will have to be updated to take a possible alias
name as well. That needs to be tracked inside QcSqliteInfo, so that
when there is a statement like "select t2.a from t1 t2" we report
the field as t1.a and not as t2.a.
The structure for storing statement information is now a C++ class.
At least initially it will not be turned into a full-fledged class
that would contain all functionality, but the code will be a mishmash
of C and C++.
- Fields in CTEs are marked as being used in subselects.
- In qc_mysqlembedded all selects must be walked if CTE is present.
- In qc_sqlite unions need special handling.
The default sql mode must now be provided explicitly when the query
classifier is setup. This is in preparation for "sql_mode" becoming
a global configuration parameter of MaxScale.
That's where it belongs as it is only the mysql client protocol that
will use it. It's a bit unfortunate that the qc test program compare
now needs to include a file from a protocol module directory, but
the fact is that the query classifier implementation and the test
programs should actually be *under* the mysql client protocol module.
