In Orion, we have a modified version of Esprima (its a bit old now,
version 2.0), that has more tolerance in it than the stock Esprima
does. It
might provide the level of recovery you are looking for.
Our recovery is done in a few ways:
1. we catch and record all exceptions, to make it far less 'throwy'
2. we perform node fill-ins when an exception would leave the AST is
bad
state
3. in some cases we rewind and try again based on state / tokens /
and
line endings
Our version can be found here:
http://git.eclipse.org/c/orion/org.eclipse.orion.client.git/tree/bundles/org.eclipse.orion.client.javascript/web/esprima/esprima.js
The tests we run on it (to give an idea of the kinds of things we can
recover from):
http://git.eclipse.org/c/orion/org.eclipse.orion.client.git/tree/bundles/org.eclipse.orion.client.javascript/web/js-tests/javascript/esprimaTolerantTests.js
And lastly, the bug for us to upgrade to the latest version of
Esprima:
https://bugs.eclipse.org/bugs/show_bug.cgi?id=473765
Michael Rennie
----- Original message -----
From: Eugene Melekhov <e...@mail.ru>
Sent by: wtp-dev-boun...@eclipse.org
To: "General discussion of project-wide or architectural issues." <
wtp-dev@eclipse.org>
Cc:
Subject: Re: [wtp-dev] JSDT Parser
Date: Sat, Mar 12, 2016 10:31 AM
Angelo,
Yes, sure, I used that option. The problem is that esprima actually
throws
a lot of syntax exceptions:
grep -c throwUnexpectedToken esprima.js
56
grep -c tolerateUnexpectedToken esprima.js
29
That means that in 2/3 error cases it throws exception and tolerate
only
1/3. Shift meanwhile doesn't throws exceptions
in "Early Error" cases like this 0=0, which is tolerated by esprima.
I
have a suspicion that most of esprima
"tolerated errors" are from that "Early Error" category, which means
that
in fact Shift behaves like esprima :-)
I thought that it would be interesting to compare shift and esprima
behavior on large enough set of erroneous code
fragments. It might happen that they are very close actually :-)
I'm not a big expert with esprima, but do you use tolerant option
like
explained at http://esprima.org/doc/ ?
2016-03-12 10:25 GMT+01:00 Eugene Melekhov <e...@mail.ru>:
Angelo,
One more thing about parser tolerance. esprima for example throws
an
exception parsing this "relatively simple"
declarations:
//function foo(a, b {}
//function foo(a, b, c,) {}
//function foo(a, b, {} {}
Yes sure, but it seems that Shift doesn't support tolerant parser
which is very required for a JS editor. See
https://github.com/shapesecurity/shift-java/issues/93
--
Eugene Melekhov
_______________________________________________
wtp-dev mailing list
wtp-dev@eclipse.org
To change your delivery options, retrieve your password, or
unsubscribe
from this list, visit
https://dev.eclipse.org/mailman/listinfo/wtp-dev
--
Eugene Melekhov
_______________________________________________
wtp-dev mailing list
wtp-dev@eclipse.org
To change your delivery options, retrieve your password, or
unsubscribe
from this list, visit
https://dev.eclipse.org/mailman/listinfo/wtp-dev
_______________________________________________
wtp-dev mailing list
wtp-dev@eclipse.org
To change your delivery options, retrieve your password, or
unsubscribe
from this list, visit
https://dev.eclipse.org/mailman/listinfo/wtp-dev
