The most important changes in version 3.1.3 are:
A number of changes makes statistics gathering more versatile:
Your rwl program can change the value of the everyuntil variable inside a control loop, and such a change will modify the expected start time of the next loop.
If you call the new nostatistics statement inside a procedure that otherwise has statistics gathered, the statistics for the current call will not be save in the various repository tables. As an example, this can be used if your procedure gets into some error situation and you therefore not want statistics to be excluded.
If you need to output values that are “computer sizes”, you can use the ‘K’ specifier. As an example, the format string “%.2KB” will output the value 1030 as 1.01KiB.
The oltp workload has had several changes including the ability to gather and output database server user and system cpu values from gv$osstat and more possibilities for emulating connection storms.
A -= assignment opertor has been added doing what it does in the C language.
The most important changes in version 3.1.2 are:
The rwloadsim tool itself and the standard generated utilities such as ociping are now available on Microsoft Windows. All details are in WINDOWS.md. Note that the oltp workload is not available and is not planned to be ported to Microsoft Windows. This is primarily due to the heavy use of Linux shell scripts.
The awrreport tool has been enhanced such that it can generate global awr reports.
If you are using immediate or embedded SQL, you can get the sqlid of such SQL by calling sql_id() with empty arguments.
In addition to the time taken to execute a procedure (stored in the ETIME columns of the runres and possibly persec repository tables) you can now also get the application time and database time. This feature is currently only available on Linux.
Note that release 3.1.1 was skipped.
Version 3.1 has particular focus on making rwloadsim more versatile as a scripting tool. The new features include ideas from SQL*Plus on how dynamic sql can be executed, and it enhances several features that were included in version 3.0. It is considered if in a future release, a separate executable, rwlscript, should be included.
There are additionally general enhancements and bug fixes.
The most important changes are:
In previous version, you could use a for loop to loop over a set of values from a low to a high value with an increment of 1. This type of loop is now called a counter loop. In version 3.1 a new loop that loops over a comma separated list of expressions has been added. The following sample for ampersand replacement shows how this new list loop can be used.
Inspired by sqlplus, you can include code such as &variable. in your directly embedded sql, which will be replaced be the actual contents of the variable during execution. As a simple example, the following will create three identically shaped tables named taba until tabc:
$ampersand:on # turn on ampersand replacement
string tablename;
for tablename := "taba", "tabb", "tabc" loop
# execute the ddl using ampersand replacement for the name of the table
create table &tablename. (a number primary key, b varchar2(10));
end loop;
A $sqllogging directive has been added allowing logging of all sql being executed, including values of any binds. Logging can be done to stdin, stderr or to a file.
There are new functions added such as floor, trunc, and the modulus operator can be used on double values.
Whenever implicit bind is in use, use can make that implicit bind imply a bindout should be used via the $bindoutname:on directive. A bind is made into a bindout when the name of the placeholder begins with a certain text string, which by default is “out”. As an example
$bindoutname:on
string outdate;
begin
:outdate := to_char(sysdate,'YYYY.MM.DD HH24:MI:SS');
end;
will return sysdate from the database into the variable outdate.
The $pre31fileassign:warn is now set implying any use of the old syntax for opening files will emit a warning. If you have not yet changed your code, you will need to do so before the upcoming release 3.1, which will have $pre31fileassign:off in effect.
The utility to create awr reports has been renamed to awrreport. Otherwise, 3.0.5 only has bug fixes and very minor changes.
Stack evaluation with mixed data types has been corrected. Previously, the entire stack would be evaluated using the same data type even if constants or variables were a different type. As an example
double d;
d := 1/2;
would assign the value 0.5 to the double variable despite the two constants are both integer. In version 3.0.4, this is now behaving is it does in e.g. C so the above is an integer division 1/2 returning 0, which is then assigned to the variable d. If expression evaluation using double is wanted, make either or both of the operators and double such as:
d := 1.0/2;
The persec repository table that contains statistics of execution for each second has two new columns, wtime and etime, which like the same columns in runres contain respectively the time spent waiting for a session and time spent executing.
A threads global attribute has been added to variables of type
integer, double and string causing the variable to be shared
among all threads.
A mutex is used to control internal data structures in rwloadsim.
The awreport utility is updated and a new utility awrdump is provided. Additionally, these two and the network test utilities are available as shell scripts in the bin directory.
The syntax for opening files and pipelines has been enhanced with new operators: >=, »=, <=, >|= and <|=. These replace the previous behavior where the contents of the concatenation being assigned to a file would be interpreted to imply open for read, append, or as a pipeline respectively. This has a security concern and is deprecated, it will be desupported in an upcoming 3.1 release, where no interpretation of characters like » or | in the file name will take place. You can enable warnings when using the current approach via the $pre31fileassign:warn directive, and it is suggested you put this directive in your .rwloadsim.rwl startup file.
In a future release case, when and switch will be keywords,
and there is therefore a warning if you use these as identifiers.
You should change your code to use different identifiers if you get this warning.
Rwloadsim will now recognize all SQL keywords and initiate scanning for SQL or PL/SQL when any of these are seen, so there is no longer a need to use an initial /* or – comment if you need SQL or PL/SQL that begins with the more rarely used keywords. You write SQL or PL/SQL, and rwloadsim sees it, knows which it is and therefore knows when to stop. It does imply, that if you have been using such a keyword as an identifier, you will now get syntax errors, and you will need to change your code to use different identifiers. As a small example, the following code that was valid in previous versions:
integer audit;
integer a,b,c;
will now generate an error about incorrect integer declaration
since audit is taken as initial keyword to start scanning
for SQL.
The three utilities ociping.rwl, connping.rwl and netthroughput.rwl are shipped as standalone, generated binaries, and they can therefore be used without a full install of the RWP*Load Simulator.
If you have a commonly used rwl script that you would like to distribute to be executed in many environments, all these would also require an installation of rwloadsim. The new feature in version 3.0.2 allows you to generate an executable that executes one specific rwl script, and which can be copied completely stand alone to any system that just has an Oracle client (or server) installation.
If you e.g. do
rwloadsim --generate=./ociping -u ociping.rwl
A binary called ociping will be created. This binary can now be copied to any system where it can be executed without an installation of rwloadsim.
This allows for much simpler distribution of (simple) rwloadsim scripts.
Version 3.0 of the RWP*Load Simulator has a number of new features that in particular make scripting, i.e. integration between bash and the Oracle database, much easier. These new features are:
In version 3, formatting based on printf and associated functions have been added. This can be done to format output to files (like printf, fprintf) as wall as to strings (like sprintf). There are some examples below. This augments the existing simple formatting with only rudimentary control via $iformat or $dformat directives.
In previous versions of rwloadsim, you were required to explicitly bind placeholders (i.e. variables such as :1 or :empno) to variables in your program, and similarly to explicitly define which variables should be receiving values from select list elements in queries. Version 3.0 can do this implicitly. Named placeholders (such as :empno) can implicitly be bound to variables of the same name (such as empno), and select list elements can implicitly have a define operation to variables.
Hence, if you execute a query like
select empno, ename, sal
from emp
where deptno=:deptno
and you have the necessary variables, empno, ename, sal and deptno declared you no longer need to provide any bind or define attributes to your sql statement.
This implicit bind and/or define can be enabled for individual sql statements using a simple syntax or generally for all sql statements using a directive. A complete example using the above query is
integer deptno := 10, empno;
double sal;
string ename;
$implicit:both # This will turn on both implicit bind and define
sql selemps
select empno, ename, sal
from emp
where deptno=:deptno;
end;
for selemps loop
printf "%d %s %.2f\n", empno, ename, sal;
end loop;
In previous versions, it was necessary to have two rwl statements associated with each sql; the first would declare the sql as a variable, and the second would execute this variable as is (for e.g. DML, PL/SQL or single row queries) or as a cursor loop for queries with an unknown number of rows. In version 3, you can now combine these into one for immediate execution.
This is done by writing code similar to a sql declaration but
using the keyword execute in stead of a name.
Also, implicit bind and define is automatically enabled.
Using this, the above example would be simplified to:
integer deptno := 10, empno;
double sal;
string ename;
for
sql execute
select empno, ename, sal from emp
where deptno=:deptno;
end
loop
printf "%d %s %.2f\n", empno, ename, sal;
end loop;
If there is no need for any sql specifications, the syntax can be simplified further using embedded sql, which here is shown with the sql being terminated by a line containing only white-space and /:
integer deptno := 10, empno;
double sal;
string ename;
for
select empno, ename, sal from emp
where deptno=:deptno
/
loop
printf "%d %s %.2f\n", empno, ename, sal;
end loop;
To prevent having passwords in clear text in either .rwl files or on the command line (from where rwloadsim does overwrite the memory), you can now use external authentication via wallets. This is achieved by declaring a database with a connect string that matches a name in the wallet but neither the username nor password attributes, or alternatively by giving empty username and password using /@name to the -l option similar to how SQL*Plus does it.