CREATE INDEX i_l_shipdate ON lineitem (l_shipdate);
CREATE INDEX i_l_suppkey_partkey ON lineitem (l_partkey, l_suppkey);
CREATE INDEX i_l_partkey ON lineitem (l_partkey);
CREATE INDEX i_l_suppkey ON lineitem (l_suppkey);
CREATE INDEX i_l_receiptdate ON lineitem (l_receiptdate);
CREATE INDEX i_l_orderkey ON lineitem (l_orderkey);
CREATE INDEX i_l_orderkey_quantity ON lineitem (l_orderkey, l_quantity);
CREATE INDEX i_c_nationkey ON customer (c_nationkey);
CREATE INDEX i_o_orderdate ON orders (o_orderdate);
CREATE INDEX i_o_custkey ON orders (o_custkey);
CREATE INDEX i_s_nationkey ON supplier (s_nationkey);
CREATE INDEX i_ps_partkey ON partsupp (ps_partkey);
CREATE INDEX i_ps_suppkey ON partsupp (ps_suppkey);
CREATE INDEX i_n_regionkey ON nation (n_regionkey);
CREATE INDEX i_l_commitdate ON lineitem (l_commitdate);

However, not every index will be used. Generating more than necessary indexes will increase the data set size. I analyzed in my settings, only 14 indexes are used (4 primary and 10 secondary):

i_l_orderkey
i_l_partkey
i_l_receiptdate
i_l_shipdate
i_l_suppkey_partkey
i_o_custkey
i_o_orderdate
i_ps_partkey
i_ps_suppkey
i_s_nationkey
pk_customer
pk_orders
pk_part
pk_supplier

We can modify “dbts/scripts/pgsql/create_indexes.sh” to reduce the number of indexes, thus reducing dataset size.

Related posts:

1. Error in rf2 of DBT3 (TPC-H)
2. Preparations to use TPCH data on Postgresql
3. Postgresql: how an index is physically created on disk

ERROR:  missing FROM-clause entry for table "tmp_orderkey1"
LINE 1: delete from lineitem where l_orderkey=tmp_orderkey1.orderkey...

ERROR:  missing FROM-clause entry for table "tmp_orderkey1"
LINE 1: delete from orders where o_orderkey=tmp_orderkey1.orderkey...

The meaning of the error message is that we must explicitly specify that we want to retrieve values stored in table “tmp_orderkey1″. This happens in TPC-H Refresh Function 2 (rf2). This function deletes records from some tables.

Let’s first review the DELETE in PostgreSQL: link.

Now let’s check out file “dbt3/scripts/pgsql/run_rf2.sh”. Towards the bottom of this file, we could see two lines of SQL commands “delete from …”. One modifies “lineitem” and the other modifies “orders”. Before the “where” of these two commands, add “using tmp_orderkey$set_num”. This will fix the problem.

Related posts:

1. Indexes created by dbt3 (TPC-H)
2. error: ‘textout’ undeclared (first use in this function)
3. Error when running DBT2: “Can’t locate Test/Parser/Dbt2.pm in @INC”

Look into file “dbt3/src/get_statement.c”. At the bottom, there is a function

Obviously, this function is supposed to trim the leading white spaces of each line. However, sometime (not always!) I get results like this:

before:  interval '1 month';
after:   interval '1 month;;

Something happened inside “strcpy”. To solve this problem, define another buffer, and replace the above Line 7 into the following Line 3 and 4:

Now the code works as expected.

Related posts:

1. Error in rf2 of DBT3 (TPC-H)
2. Code to compute SHA-1 hash for a file
3. How to implement file upload with CGIC

int mode=1;
module_param(mode, int, 444);

System call module_param creates an file in the sysfs. The pathname might be: /sys/module/modname/parameters/mode. Here, modename is the registered name of your module. By using “444″ in module_param, we make this file writable. You can use “sudo echo … > /sys/…/mode” to change the content of this file.

When the file is changed, your module will also see the change. However, sysfs does NOT tell the module about this change. Therefore, in your code, you may have to proactively check its value. The code could be like this:

extern int mode;
static int old_mode=mode;
void thread_run(){ //a repeatedly running thread
       sleep(1);
       if (mode != old_mode){
               //code to handle mode change
               ...
               old_mode=mode;
       }
}

Related posts:

1. Trouble shooting for writing your first Linux module on Ubuntu
2. Write to a file in PHP
3. Location of the kernel symbol table on Fedora Linux

c0100000 T startup_32
c0100000 T _text
c0100079 t bad_subarch
c0100079 W lguest_entry
c0100079 W xen_entry
c010007b t default_entry
c0101000 T wakeup_pmode_return
c010104c t bogus_magic
c010104e t save_registers
c010109d t restore_registers

To understand the meaning of letters in the middle column, such as “T”, “t”, and “W”, you need to check out the manpage of “nm”.

Related posts:

1. List the content of Kernel Symbol Table
2. Linux kernel messages are not printed on the console
3. The document.location in JavaScript

aclocal
autoreconf --install

In the process, you may need to install packages “automake” and “libtool”.

The run

./configure --with-postgresql

or “–with-mysql” or “–with-sapdb”. If you do not give the “with” parameter, the generated Makefile will not work correctly.

Related posts:

1. Error when compiling postgresql
2. DBT3 (TPC-H) user manual
3. Error when running DBT2: “Can’t locate Test/Parser/Dbt2.pm in @INC”

Related posts:

1. Some notes about compiling DBT3 (TPC-H)

(1) CreateQueryDesc (tcop/pquery.c)
(2) ExecutorStart (executor/execMain.c)
    (2.1) CreateExecutorState (executor/execUtils.c)
    (2.2) ExecInitNode (executor/execProcnode.c)
        (2.2.1) CreateExprContext (executor/execUtils.c)
        (2.2.2) ExecInitExpr (executor/execQual.c)
(3) ExecutorRun (executor/execMain.c)
    (3.1) ExecProcNode (executor/execProcnode.c)
        (3.1.1) ExecEvalExpr (include/executor/executor.h)
        (3.1.2) ResetExprContext (include/executor/executor.h)
(4) ExecutorEnd (executor/execMain.c)
    (4.1) ExecEndNode (executor/execProcnode.c)
    (4.2) FreeExecutorState (executor/execUtils.c)
(5) FreeQueryDesc (tcop/pquery.c)

“CreateQueryDesc” is will be invoked through a series of calls following

exec_simple_query (tcop/postgres.c)
|->1. pg_analyze_and_rewrite (tcop/postgres.c)
|->2. pg_plan_queries (tcop/postgres.c)
|->3. PortalStart (tcop/pquery.c)
|-----> ...

“A portal is an abstraction which represents the execution state of a running or runnable query.” — “include/utils/portal.h”. In the comments of the header file, we can see the strategies used by different types of queries.

Related posts:

1. A TPC-H Q21 query plan
2. Error when compiling postgresql
3. How to debug postgresql during initdb

:res+10

To decrease by 10, type command

:res-10

2. Change the width (number of columns) of a split window
To increase by 10, type command

:vertical res+10

To decrease by 10, type command

:vertical res-10

Please note that some people use “+W” shortcuts to do the same job, but I am having trouble when using these shortcuts. So I just stick to changing the size by typing the commands.

Related posts:

1. Adjust Partitions in Ubuntu on the fly
2. Split one text file into multiple ones, with Shell script
3. vim NERDTree cheat sheet

select
        s_name,
        count(*) as numwait
from
        supplier,
        lineitem l1,
        orders,
        nation
where
        s_suppkey = l1.l_suppkey
        and o_orderkey = l1.l_orderkey
        and o_orderstatus = 'F'
        and l1.l_receiptdate > l1.l_commitdate
        and exists (
                select
                        *
                from
                        lineitem l2
                where
                        l2.l_orderkey = l1.l_orderkey
                        and l2.l_suppkey <> l1.l_suppkey
        )
        and not exists (
                select
                        *
                from
                        lineitem l3
                where
                        l3.l_orderkey = l1.l_orderkey
                        and l3.l_suppkey <> l1.l_suppkey
                        and l3.l_receiptdate > l3.l_commitdate
        )
        and s_nationkey = n_nationkey
        and n_name = 'BRAZIL'
group by
        s_name
order by
        numwait desc,
        s_name;

Query plan

Related posts:

1. Control flow of a query in PostgreSQL