[Pipet Devel] Peep

[J.W. Bizzaro]

deanne at density.biop.umich.edu wrote:
> 
> If you put in something that's not in the interpretation list
> (do_<blah>) you just get a "Unknown syntax" error. If you hit return for a
> blank line, it repeats the last command, though I'm sure I can override
> that.

I'd like it to work as much like the Unix shell/CLI as possible.  So, hitting
return on a blank line should give you another prompt.

> <peep> nodeset
> 
> Enter the number of  the new node: 1
> Enter new node type: type1
> Enter the remote IP or name for this node: 121.00.00.00
> Node number:  1     Node Type:  type1     Remote IP:  121.00.00.00

Interesting.  I was thinking about a non-interactive, all-in-one command
like...

  nodeset number=1 type=1 ip=121.0.0.0

But command-lines like this are hard to remember.  How about we combine the
two: If the user enters an incomplete command, Peep asks for the extra values?

> <peep> nodeset
> 
> Enter the number of  the new node: 2

Okay, maybe this is just a test/example, but node numbering should be
automatic...probably handled by the UI (not the DL).

> Enter new node type: type2
> Enter the remote IP or name for this node: 121.00.00.01

These two are really for specifying a BRAND NEW node, and I'm not sure how
much or if that will be needed.  Piper should know what is on the local
filesystem and any remote systems you're connected to.  The user should be
able to type a "list" command to see all basic nodes available and all
pre-prepared networks.

I think the only time the user will need to enter an IP is to link with a
Piper on another, known system.

> <peep> nodelist
> 
> Current Configured Nodes
> ------------------------
> Node #    Node Type      IP address
> =========================================
>    1      type1       121.00.00.00
>    2      type2       121.00.00.01
>    3      type3       12.1.11.0

Yeah, this is what I was thinking about.  It's something like the ls command,
aint it?

Again, about node numbering: Actual node ID's will be pretty long.  Brad had a
good idea to make the path the ID, something like...

    ID=unix/commands/ls

I think we could make it more unique by using a full URI like...

    ID=piper://bioinformatics.org:1000/unix/commands/ls

That's definitely unique (no duplicate nodes on the net).

But then it would be kind of silly to have to type that out on the
command-line.  Imagine...

    connect locus=piper://bioinformatics.org:1000/unix/commands/ls:out1pipe
locus=piper://bioinformatics.org:1000/viewers/text:in1pipe

If the UI would simply assign a number (dynamically -- if a locus is removed,
everything is renumbered) to each locus:pipe in the network, it'd be a lot
easier on the user.  So, if...

    piper://bioinformatics.org:1000/unix/commands/ls = 1
    piper://bioinformatics.org:1000/unix/commands/ls = 2

and

    out1pipe = C
    in1pipe = A    (attached to different loci, actually)

the user can type...

    connect 1:C 2:A

Simple.

Peep may be small now, but I expect it to gain A LOT of functionality.  Still,
it's meant to be small and light-weight compared to Pied.

Cheers.
Jeff
-- 
                      +----------------------------------+
                      |           J.W. Bizzaro           |
                      |                                  |
                      | http://bioinformatics.org/~jeff/ |
                      |                                  |
                      |        BIOINFORMATICS.ORG        |
                      |           The Open Lab           |
                      |                                  |
                      |    http://bioinformatics.org/    |
                      +----------------------------------+