Replacing the query optimizer is a long-term work, I think it is not within the scope of this discuss.
Our main concern here is how to ensure normal compilation when the address of the maven repo is changed. Here I still suggest to turn this variable address into an environment variable, and update our compilation and deployment guide after the address changes, so that users can solve the problem by adjusting the compilation parameters without modifying the code. -- 此致！Best Regards 陈明雨 Mingyu Chen Email: chenmin...@apache.org 在 2020-10-28 16:48:36，"ling miao" <lingm...@apache.org> 写道： >Hi Duo, > >I found another dependency package with similar functions, let's see if I >can replace it. > >Regarding the question of Calcite, we have previously investigated Calcite >for Doris to optimize the query framework. >After the evaluation, it is found that the access cost may be similar to >the reconstruction of the entire query framework , so it is a long-term >work in the future. > >Ling Miao > > >张铎(Duo Zhang) <palomino...@gmail.com> 于2020年10月27日周二 下午9:53写道： > >> Could be a long term solution? Is there any other reason which prevents us >> using Calcite? >> >> ling miao <lingm...@apache.org> 于2020年10月27日周二 下午8:02写道： >> >> > Hi Duo, >> > >> > Using calcite directly will definitely not work, which means that the >> query >> > analysis + planning are rewritten. >> > But maybe we can use other parser tools to replace it, or something >> > similar. >> > However, it is estimated that the amount of change for this solution. >> > ...(｡•ˇ‸ˇ•｡) >> > ... >> > >> > Ling Miao >> > >> > 张铎(Duo Zhang) <palomino...@gmail.com> 于2020年10月27日周二 下午6:34写道： >> > >> > > I think this is used to generate the SQL parser? Is it possible to use >> > > Calcite directly? >> > > >> > > ling miao <lingm...@apache.org> 于2020年10月26日周一 下午7:38写道： >> > > >> > > > Hi, >> > > > >> > > > Currently, some fe dependencies that we rely on are not in the maven >> > > repo. >> > > > For example, cup-maven-plugin and java-cup are located at the >> Cloudera >> > > > 3rd-P. >> > > > This will lead to the risk that our code will not be compiled when >> the >> > > > third-party repo changes. >> > > > >> > > > I have also tried whether it is possible to replace these >> dependencies >> > > > with other packages or versions, but unfortunately the only thing we >> > can >> > > > use at the moment is this. >> > > > >> > > > So I propose whether we can maintain these jar packages in our own >> repo >> > > to >> > > > ensure that they can be found every time we compile? >> > > > >> > > > Or do you have any better solutions? >> > > > >> > > > Ling Miao >> > > > >> > > >> > >>
