[Mulgara-dev] TqlSession changes

Paul Gearon gearon at ieee.org
Tue Nov 4 04:29:22 UTC 2008


On Mon, Nov 3, 2008 at 7:33 PM, Brian Sletten <brian at bosatsu.net> wrote:
> Hi, Alex, thanks for checking this out. I will look into the problems you
> are having, but I wanted to address a few points.
>
>> 1. In production, I never use the default "server1" server name.  In addition to
>> specifying a hostname for querying graphs for auto-completion, it would be nice
>> to override the default server name and RMI registry port.
>
> This is why I am going to be focusing on a new shell that allows you to
> connect to servers like:
> connect <rmi://localhost/server1>;

Yes please. Also, please don't intercept these commands.... I'm hoping
you'll put them into the tql grammar like everything else. (If not,
then let me know and I'll do it for you, and let you hook into it).

<snip/>
>> 2. No messages are actually flushed to the console until a query is successfully
>> executed.  For instance, I start the shell and type "help;" and hit enter -
>> nothing happens.  I type in a query and hit enter, and then the help message is
>> displayed, followed by the query results.  The same thing happens when there is
>> an error parsing or executing a command - the error message is not displayed
>> until a query actually completes.
>
> I'll see if I can fix this in the short-term, but I will certainly make the
> new shell work like this.

+1 from me. This is unbelievably painful!

For instance, I initially thought that you must have done the
"connect" command, so I issued it... only to get no response. In fact,
everything was giving me no response ("Hmmm," I thought. "I *suppose*
it's working...."). Then when I did a select I got back several
answers, including "Error parsing the query". It took me a moment to
realize that the "query" it was complaining about was not my "select"
query, but the unsuccessful "connect".

Another thing to note was that in a separate session I set the timer
on and got no response on loading a file. I suppose that was because I
didn't do a "select" in that session.

As I said.... painful.

>> 3. I can get into a weird state where I can't exit the shell with the "quit"
>> command.  The most reliable way to reproduce this is to open a new shell, type
>> in a query with a syntax error and hit enter.  Entering "quit;" at the next
>> command prompt does not do anything.  There seems to be something screwy
>> going on with the command splitting or something like that.
>
> I think right now CTRL-D is the only thing that shuts it down, but I will
> make sure quit works as well.

"quit" has worked for me every time so far, but I haven't got into
that "weird" state yet.

> All in all though, the Swing-less command shell is a much-anticipated
> improvement.  Thanks!

+1  :-)

> Sure. I think people will like it. I've waited years for this capability!
> I plan to add typical shell "!" commands (last select, last load, last
> connect, etc.), history management, etc.
> If anyone else thinks of desired features, let me know.

Hey, *I* wanted it too. I used home-grown command line tools myself
(look in the "tools" directory - though my own stuff was a little more
developed than these). There was NOTHING redeeming about the Swing
client, except that it had history. And "history" was the reason it
was implemented in Swing. JLine has removed the one obstacle we had in
creating a command line client.

Thanks Brian. It's much appreciated.

Regards,
Paul



More information about the Mulgara-dev mailing list