incubator-general mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Kasper Sørensen <>
Subject RE: [PROPOSAL] Optiq
Date Thu, 01 May 2014 20:19:14 GMT
I am certainly not questioning the power of Optiq. Just noting that I think it has a lot of
similarities, and obviously there are differences. But even in your list I find a lot of the
points to be close or similar. I would hate to start a war over words in this thread, since
I actually only wanted to point to a related project. But to your points:

- MetaModel also provides SQL support, including nested/sub-queries. Not correlated sub-queries
though, which is a difference.

- SQL parsing is a goal. Not exactly sure what you mean by "transformations"; I think it's
either User Defined Functions (UDFs) or transformations of the query itself to fit with a
particular backend. UDFs are currently not supported no, but we do quite a lot of tricks to
transform and optimize the query plan based on the backing store.

- Sounds like Optiq is ahead of MM in terms of query transformations.

- What do you mean when you say query execution is not a central goal of Optiq? What would
you otherwise be needing your query for?

- Can you explain or link to more information about the type inference you mention?

Best regards,

From: Ted Dunning []
Sent: 01 May 2014 09:14
Subject: Re: [PROPOSAL] Optiq

I think that there is a huge difference between Metamodel and Optiq.

In particular:

- Optiq provides real SQL including nested queries, correlated sub-queries
and so on

- Metamodel uses a fluent Java API ... SQL parsing and transformation
doesn't appear to be a goal

- Optiq provides highly advanced query transformations including
decorrelations based on estimated execution costs.

- Metamodel appears to provide no significant query transformations

- Optiq only provides query execution as a by-product for testing

- Metamodel has query execution as a central goal

- Optiq provides a form of type inferencing for SQL queries.  This is
unique to Optiq as far as I know.

On Thu, May 1, 2014 at 8:57 AM, Kasper Sørensen <> wrote:

> I see a lot of conceptual similarity between Optiq and the Apache
> MetaModel (incubator) project [1]. Maybe something can be done to align the
> two projects, so that we avoid having two incubating projects that do
> basically the same thing?
> Or maybe there's some glaring difference that I am missing? At least it
> seems to me both to be projects that try to provide uniform querying
> capabilities to a wide array of data backends. Both project also favor a
> type-safe Java querying API instead of a String/SQL oriented query API.
> Regards,
> Kasper Sørensen
> [1]
> ________________________________________
> From: Ashutosh Chauhan []
> Sent: 01 May 2014 00:21
> To:
> Subject: [PROPOSAL] Optiq
> I would like to propose Optiq as an Apache Incubator project.  I have
> posted the proposal to and
> posted the text of the proposal below.
> Ashutosh.
> = Optiq =
> == Abstract ==
> Optiq is a framework that allows efficient translation of queries involving
> heterogeneous and federated data.
> == Proposal ==
> Optiq is a highly customizable engine for parsing and planning queries on
> data in a wide variety of formats. It allows database-like access, and in
> particular a SQL interface and advanced query optimization, for data not
> residing in a traditional database.
> == Background ==
> Databases were traditionally engineered in a monolithic stack, providing a
> data storage format, data processing algorithms, query parser, query
> planner, built-in functions, metadata repository and connectivity layer.
> They innovate in some areas but rarely in all.
> Modern data management systems are decomposing that stack into separate
> components, separating data, processing engine, metadata, and query
> language support. They are highly heterogeneous, with data in multiple
> locations and formats, caching and redundant data, different workloads, and
> processing occurring in different engines.
> Query planning (sometimes called query optimization) has always been a key
> function of a DBMS, because it allows the implementors to introduce new
> query-processing algorithms, and allows data administrators to re-organize
> the data without affecting applications built on that data. In a
> componentized system, the query planner integrates the components (data
> formats, engines, algorithms) without introducing unncessary coupling or
> performance tradeoffs.
> But building a query planner is hard; many systems muddle along without a
> planner, and indeed a SQL interface, until the demand from their customers
> is overwhelming.
> There is an opportunity to make this process more efficient by creating a
> re-usable framework.
> == Rationale ==
> Optiq allows database-like access, and in particular a SQL interface and
> advanced query optimization, for data not residing in a traditional
> database. It is complementary to many current Hadoop and NoSQL systems,
> which have innovative and performant storage and runtime systems but lack a
> SQL interface and intelligent query translation.
> Optiq is already in use by several projects, including Apache Drill, Apache
> Hive and Cascading Lingual, and commercial products.
> Optiq's architecture consists of:
> An extensible relational algebra.
> SPIs (service-provider interfaces) for metadata (schemas and tables),
> planner rules, statistics, cost-estimates, user-defined functions.
> Built-in sets of rules for logical transformations and common data-sources.
> Two query planning engines driven by rules, statistics, etc. One engine is
> cost-based, the other rule-based.
> Optional SQL parser, validator and translator to relational algebra.
> Optional JDBC driver.
> == Initial Goals ==
> The initial goals are be to move the existing codebase to Apache and
> integrate with the Apache development process. Once this is accomplished,
> we plan for incremental development and releases that follow the Apache
> guidelines.
> As we move the code into the org.apache namespace, we will restructure
> components as necessary to allow clients to use just the components of
> Optiq that they need.
> A version 1.0 release, including pre-built binaries, will foster wider
> adoption.
> == Current Status ==
> Optiq has had over a dozen minor releases over the last 18 months. Its core
> SQL parser and validator, and its planning engine and core rules, are
> mature and robust and are the basis for several production systems; but
> other components and SPIs are still undergoing rapid evolution.
> === Meritocracy ===
> We plan to invest in supporting a meritocracy. We will discuss the
> requirements in an open forum. We encourage the companies and projects
> using Optiq to discuss their requirements in an open forum and to
> participate in development. We will encourage and monitor community
> participation so that privileges can be extended to those that contribute.
> Optiq's pluggable architecture encourages developers to contribute
> extensions such as adapters for data sources, new planning rules, and
> better statistics and cost-estimation functions. We look forward to
> fostering a rich ecosystem of extensions.
> === Community ===
> Building a data management system requires a high degree of technical
> skill, and correspondingly, the community of developers directly using
> Optiq is potentially fairly small, albeit highly technical and engaged. But
> we also expect engagement from members of the communities of projects that
> use Optiq, such as Drill and Hive. And we intend to structure Optiq so that
> it can be used for lighter weight applications, such as providing a SQL and
> JDBC interface to a NoSQL system.
> === Core Developers ===
> The developers on the initial committers list are all experienced open
> source developers, and are actively using Optiq in their projects.
>  * Julian Hyde is lead developer of Mondrian, an open source OLAP engine,
> and an Apache Drill committer.
>  * Chris Wensel is lead developer of Cascading, and of Lingual, the SQL
> interface to Cascading built using Optiq.
>  * Jacques Nadeau is lead developer of Apache Drill, which uses Optiq.
> In addition, there are several regular contributors whom we hope will
> graduate to committers during the incubation process.
> We realize that additional employer diversity is needed, and we will work
> aggressively to recruit developers from additional companies.
> === Alignment ===
> Apache, and in particular the ecosystem surrounding Hadoop, contains
> several projects for building data management systems that leverage each
> other's capabilities. Optiq is a natural fit for that ecosystem, and will
> help foster projects meeting new challenges.
> Optiq is already used by Apache Hive and Apache Drill; Optiq embeds Apache
> Spark as an optional engine; we are in discussion with Apache Phoenix about
> integrating JDBC and query planning.
> == Known Risks ==
> === Orphaned Products ===
> Optiq is already a key component in three independent projects, each backed
> by a different company, so the risk of being orphaned is relatively low. We
> plan to mitigate this risk by recruiting additional committers, and
> promoting Optiq's adoption as a framework by other projects.
> === Inexperience with Open Source ===
> The initial committers are all Apache members, some of whom have several
> years in the Apache Hadoop community. The founder of the project, Julian
> Hyde, has been a founder and key developer in open source projects for over
> ten years.
> === Homogenous Developers ===
> The initial committers are employed by a number of companies, including
> Concurrent, Hortonworks, MapR Technologies and We are
> committed to recruiting additional committers from outside these companies.
> === Reliance on Salaried Developers ===
> Like most open source projects, Optiq receives substantial support from
> salaried developers. This is to be expected given that it is a highly
> technical framework. However, they are all passionate about the project,
> and we are confident that the project will continue even if no salaried
> developers contribute to the project. As a framework, the project
> encourages the involvement of members of other projects, and of academic
> researchers. We are committed to recruiting additional committers including
> non-salaried developers.
> === Relationships with Other Apache Products ===
> As mentioned in the Alignment section, Optiq is being used by Apache Hive
> and Apache Drill, and has adapters for Apache Phoenix and Apache Spark.
> Optiq often operates on data in a Hadoop environment, so collaboration with
> other Hadoop projects is desirable and highly likely.
> === An Excessive Fascination with the Apache Brand ===
> Optiq solves a real problem, as evidenced by its take-up by other projects.
> This proposal is not for the purpose of generating publicity. Rather, the
> primary benefits to joining Apache are those outlined in the Rationale
> section.
> == Documentation ==
> Additional documentation for Optiq may be found on its github site:
>  * [[|Overview]]
>  * [[
>  * [[|HOWTO]]
>  * [[
> ]]
> === Presentation: ===
>  *[[
> SQL on Big Data using Optiq]]
> == Initial Source ==
> The initial code codebase resides in three projects, all hosted on github:
>  *
>  *
>  *
> === Source and Intellectual Property Submission Plan ===
> The initial codebase is already distributed under the Apache 2.0 License.
> The owners of the IP have indicated willingness to sign the SGA.
> === External Dependencies ===
> Optiq and Linq4j have the following external dependencies.
>  * Java 1.6, 1.7 or 1.8
>  * Apache Maven, Commons
>  * JavaCC (BSD license)
>  * Sqlline 1.1.6 (BSD license)
>  * Junit 4.11 (EPL)
>  * Janino (BSD license)
>  * Guava (Apache 2.0 license)
>  * Eigenbase-resgen, eigenbase-xom, eigenbase-properties (Apache 2.0
> license)
> Some of Optiq's adapters (optiq-csv, optiq-mongodb, optiq-spark,
> optiq-splunk) are currently developed alongside core Optiq, and have the
> following additional dependencies:
>  * Open CSV 2.3 (Apache 2.0 license)
>  * Apache Incubator Spark
>  * Mongo Java driver (Apache 2.0 license)
> Upon acceptance to the incubator, we would begin a thorough analysis of all
> transitive dependencies to verify this information and introduce license
> checking into the build and release process by integrating with Apache Rat.
> === Cryptography ===
> Optiq will eventually support encryption on the wire. This is not one of
> the initial goals, and we do not expect Optiq to be a controlled export
> item due to the use of encryption.
> == Required Resources ==
> === Mailing Lists ===
>  *
>  * (will be migrated from
>  *
> === Source control ===
> The Optiq team would like to use git for source control, due to our current
> use of git/github. We request a writeable git repo git://
>, and mirroring to be set up to github
> through INFRA.
> === Issue Tracking ===
> Optiq currently uses the github issue tracking system associated with its
> github repo: We will migrate
> to
> the Apache JIRA:
> == Initial Committers ==
>  * Julian Hyde (jhyde at apache dot org)
>  * Jacques Nadeau (jacques at apache dot org)
>  * James R. Taylor (jamestaylor at apache dot org)
>  * Chris Wensel (cwensel at apache dot org)
> === Affiliations ===
> The initial committers are employees of Concurrent, Hortonworks, MapR and
>  * Julian Hyde (Hortonworks)
>  * Jacques Nadeau (MapR Technologies)
>  * James R. Taylor (
>  * Chris Wensel (Concurrent)
> == Sponsors ==
> === Champion ===
>  * Ashutosh Chauhan (hashutosh at apache dot org)
> === Nominated Mentors ===
>  * Ted Dunning (tdunning at apache dot org) - Chief Application Architect
> at MapR Technologies; committer for Lucene, Mahout and ZooKeeper.
>  * Alan Gates (gates at apache dot org) - Architect at Hortonworks;
> committer for Pig, Hive and others.
>  * Steven Noels (stevenn at apache dot org) - Chief Technical Officer at
> NGDATA; committer for Cocoon and Forrest, mentor for Phoenix.
> === Sponsoring Entity ===
> The Apache Incubator.
> ---------------------------------------------------------------------
> To unsubscribe, e-mail:
> For additional commands, e-mail:

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message