calcite-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Julian Hyde <>
Subject Re: Vagrant/docker dev VM
Date Tue, 10 Feb 2015 20:44:41 GMT
Yes, sorry, keep meaning to try the VM. Been working a lot on my 4GB MacBook Air recently.

And, sorry I broke Mongo. I will fix it.

An assert that goes off due to user code should be changed to an explicit "if ... throw" (if
not a real user error). I'll look at how the Mongo code was using the API wrong, and change
the assert to something helpful & informative. Hopefully we can make it so that existing
code will just work.

I hear what you are saying about major versions and compatibility. However, we have a development
style where we produce a feature, let people kick the tires, and fix/adapt it over the following
few months. Over those months we will have made several releases, so the code will be "out
there". I think it would be ridiculous to release a major version more than once a year. It's
more important that we think about our users' convenience, and the burden on Calcite of maintaining
absolute compatibility, and find an acceptable balance between the two.

For example, collations never really worked, so any code that used them was by definition
experimental. We should ensure that code continues to compile, but we should not slavishly
preserve the old functionality.

I have found the '@Deprecated // to be removed before x.x' pattern useful. I can easily search
for uses of java.lang.Deprecated in my IDE and remove old APIs just before a given release.


> On Feb 10, 2015, at 2:12 AM, Vladimir Sitnikov <> wrote:
> Julian,
> Please go ahead and try the VM.
> It looks like recent trait changes defeated MongoAdapter (see my
> comment in
> The assert line is "assert traits.allSimple();"
> Should we try to use helpful messages for the assertions? Having exact
> contents of "traits" in the assert message would help to analyze
> issues from Drill/Hive/etc.
> More interesting question: should we treat this issue as a "backward
> incompatible change"?
> If no objections appear, I would add notes from
> to the
> That would close CALCITE-507.
> Vladimir

View raw message