[Linuxsampler-devel] LSCP shell

[Christian S. <sch...@li...>]

Hi!

I planned to write a shell command line application for LSCP. For this I 
already played around with Bison's auto generated parser tables. As a first 
result of these experiments, the sampler is now providing more useful error 
messages on LSCP syntax errors. Before, if you i.e. sent "GET S" (obviously 
wrong) it was spitting out something like:

	syntax error, unexpected '\r', expecting 'E' or 'T'

now it is also spitting out this:

	Context: "GET S^..." -> Should be: SEND_EFFECT_CHAINS | SERVER | STREAMS

For the LSCP shell I planned some convenient usability features for dealing 
with LSCP from the command line, either by typing and/or "running" LSCP script 
files:

	1. Auto completion: when tab is pressed, or when return is pressed and 
	   the command line not fully completed yet and only a unique solution 
	   exists to complete the command line.

	2. Auto correction: if the user mistyped something and it is obvious what 
	   he "meant", then the shell will automatically correct it for the user 
	   while typing.

	3. Colored output. Probably even enabled by default.

	4. Much more user friendly error messages on syntax errors, spread over 
	   several lines, with color and indicating semi-graphically the position 
	   of the issue. Maybe similar to clang compiler syntax error messages.

	5. Visually aided suggestion in real-time while typing. The shell could 
	   display the possibilities to complete the current command line in the 
	   same line in thin font. Once the typed command line is at a point 
	   where there is only one solution to complete it, the line could i.e. 
	   turn green.

Probably a bunch of more features, which make sense to be implemented as part 
of a shell. For example relative path arguments (see bug #154), that is 
relative to the working directory where the shell was called from. Or 
conveniently saving and restoring a sampler session with one command to/from 
the user's home directory. Or even variables to be reused in subsequent lines.

I am currently unresolved yet though where to position the shell:

a) on server side with the shell application just being a light-weight 
application that only transmits characters and prints out content in color to 
the console.

	Advantage: shell application would be easy to maintain and quite 	
	independent of the sampler's version being connected to. I.e. the same 
	shell application could deal with different LSCP versions without 
	problems.

	Disadvantage: typing would be more sluggish on remote network 
	connections, since every single character typed would be transmitted, and 
	result of its processing sent back.

Or

b) a full fledged autonomous shell application on client computer side, which 
implements the complete shell on its own, with all the parser work, error 
recovery and so on.

	Advantage: allows very quick and responsive typing.

	Disadvantage: the shell would assume/support exactly one LSCP version 
	*or* it would need to connect to the server, ask for the LSCP version 
	(GET SERVER INFO), but in that case we might need to maintain all LSCP 	
	versions inside the shell application, i.e. separate lscp.y files for 
	each LSCP version in the shell's source files.

Just thoughts for now. Comments, additions appreciated.

CU
Christian