Test/improve/enhance the command language

[kilian]

The command language is still incomplete or the implementation doesn't work (options, blocks etc.) and has to be complete for the beta release. Fine-tuning should be done for the public 2.0 version.

[kilian]

Changeset [1289] resolves part of this.

The syntax of variables is now (hopefully) reaching a stable state. For commands that take options, all options that make sense (essentially, those that set variables or other special objects such as cuts) are now allowed and should work.

The remaining tasks include eliminating redundancies (such as the list command, which could be replaced by another version of show), reconsidering the loop command, probably adding a conditional construct, and implementing external I/O (including a log file).

[kilian]

Loops are done now [1299]. I had to modify some of the code corresponding to variables, and this has broken SLHA reading. To be fixed asap.

[kilian]

The problems with MSSM handling (not a SLHA problem, but name clash A0=parameter vs. A0=particle) are solved in Changeset [1301]. Variable and loop handling seems to work as it should now.

[kilian]

Changeset [1302] implements conditionals. The missing pieces are now:

• Elementary file management (logfile and such)

• Some user options, e.g., LHAPDF sets

• clear/delete commands, e.g., for histograms

[kilian]

'clear' and some other additions are now present. The logfile works, at least for the main program and common cases. (Logging/redirecting output of embedded shell commands is not implemented, however.) I'm not yet satisfied with the file management, but the remaining changes are minimal. So, we're almost feature complete.

[Juergen Reuter]

WK now added the <expect> syntax in the command language. I proposed a first suggestion for the name of the command language, which has to be discussed. Now we have basically only testing and minor improvements/enhancements. I move the ticket to the next milestone and reduce the priority.

[Juergen Reuter]

I list a couple of problems here:

01) SINDARIN does not allow a minus sign after an '=': real foo=-2 triggers a fatal. 02) The following completey screws the user

    real foo = 0.01
    real a =   2.0
    int foo = 4
    a = a + foo

as the real foo supersedes the integer, altough being defined later.

03) Error messages are really cryptical.

[Juergen Reuter]

04) Add the possibility of having complex input parameters for models or at least derived parameters. E.g. WHIZARD could split them automatically into their real and complex parts.

[Juergen Reuter]

No, there are some actual important tasks, after they are done by Wolfgang, we can move it.

[kilian]

Long overdue: cleaned up event-file specifications in r2225.

Example:

simulate (foo, bar) {
  $sample = "foobar"
  sample_format = ascii, hepevt, lhef
}

This will write files "foobar.evt", "foobar.hepevt", "foobar.lhef".

The specification of $sample can be omitted, it defaults to the first process in the list (here: "foo").

The extensions can be redefined, if necessary, as $extension_ascii, $extension_hepmc etc.

Exception: The handling of raw event files (default: on) is controlled by ?write_raw and ?read_raw, as before.

[kilian]

Change in r2226: The problem 01) above is now solved by assigning '=' and '-' to different character classes. This has some important consequences:

• The token -> is replaced by =>. This affects process definitions, beam definitions, and parameter scans.
• The token /= is replaced by <> (as we agreed during the last WHIZARD meeting).
• The token /~ is replaced by <>~.
• For clarity, the token == is replaced by ==~.

I have modified the test files (make check) accordingly. However, all input files not in the test suite have to be modified as well. Sorry for that.

[kilian]

r2246 introduces 'sprintf' and 'printf', essentially interfaces to the corresponding C funtions (stdio.h).

Usage:

$string = "foo"
printf $string

replacing 'echo';

printf "In model %s, the W mass is %g." ($model_name, mW)

showing the syntax for formatting variables. (Currently, only variables, no expressions in the argument!)

Similarly

$string = "bar"
$foobar = sprintf "foo%s" ($string)

yields 'foobar'.

The 'echo' command is redundant now and should be eliminated (just replace by 'printf').

Restrictions:

• Only strings, integers, and reals are supported (will add complex and aliases later), logicals as strings.
• The format string is parsed and checked against the arguments (no format string exploits via WHIZARD!)
• Only standard output channel (with logfile copy) supported so far

[kilian]

The problem 02) is now solved in r2272. It adds the requirement for declaring all user variables and streamlines their usage. The other important change is reqiring a '=' sign for the loop variable assignment in scans.

[Juergen Reuter]

This is basically done, except for the complex variables. WIll see whether we need them or not. Closing this for now.

[Juergen Reuter]

Milestone v2.0.0final deleted