Re: [orca-list] Creating new scripts

[Willie Walker <walker willie gmail com>]

Hi Michael:

> I am wondering what is the best way to start writing a new script? I have read a number of docs suggested 
> by Willie Walker, they were useful. My question is more, do I copy the default script and modify the copy 
> or is there a way I can tell a completely new script to subclass/extend the default script?

Subclassing default.py or one of its subclasses would be the way to go.  Copying default.py just duplicates 
code and sets up a maintenance mess.  :-)  You might take a look under the src/orca/scripts/apps directory 
for some real life examples.

> Also is it a bit of guess work on choosing the best event to listen to? This is related to the bug I 
> mention in my last message, I have set myself the challenge to get orca to speak the label when the content 
> changes but at-spi reports three events which look suitable (one for the text being inserted, one for text 
> being deleted and one for visible data changing, I guess its probably the visible data changed one).

Yeah - there is a bit of guesswork and sleuthing that needs to be done for choosing the best event (or set of 
events) to listen to.  The way we typically do this is by turning the debug level on to maximum output and 
then scouring the debug log.  If the event model of AT-SPI were tightly spec'd, we might be able to get away 
with a state machine.  However, it is not tightly spec'd.  So, we tend to look for sentinel events, and we 
try to look for ones that are more specific than general.

For your specific example, the visible data changing one tends to be way too generic, so the text change 
event route is probably better.  To figure out which one to use, you can poke at the object in the event 
handler and check to see if its state is consistent with what is on the screen - there's no specified order 
for events, so you just need to play.

Note also that we generally try to keep event handling somewhat stateless in that we try to keep as little 
stored history inside Orca as possible. There is some stateful stuff in there (e.g., what was the last object 
with focus?), but we do try to keep the need for that down to a minimum.

Will