diff --git a/proposals/1_BASE64_ENCODING.md b/proposals/1_BASE64_ENCODING.md new file mode 100644 index 000000000..be88f8639 --- /dev/null +++ b/proposals/1_BASE64_ENCODING.md @@ -0,0 +1,44 @@ + +--- +Author: Julien Le Dem +Created: 2025-Aug-7 +Name: add BASE64 compression +Issue: https://github.com/apache/parquet-format/issues/NNN +Status: ARCHIVED +Reason: Did not compress +--- + +# Proposal + +**NOTE**: This is an example proposal for use as a template + +## Description +Add Base64 to compression algorithms. +This is not backwards compatible as a new compression alg. + +## Spec + +See [BASE64 spec]. + +## Evaluation + +After trying out in the java implementation, file size doubled on average. +See prototype [here](github.com/julienledem/mypoc) + diff --git a/proposals/README.md b/proposals/README.md new file mode 100644 index 000000000..fa7aa2f31 --- /dev/null +++ b/proposals/README.md @@ -0,0 +1,79 @@ + +# Proposals + +This proposal process is intended for significantly impactful additions to the Parquet spec. The goal is to facilitate those projects and help them being contributed to Parquet. +For example, changes that are not forward compatible like adding a new encoding or a new compression algorithm (older implementations can not read new files). [see guidelines for more details](https://github.com/apache/parquet-format/blob/master/CONTRIBUTING.md#general-guidelinespreferences-on-additions) +This gives better visibility to those projects which require coordination in several implementations. +Bug fixes, code only features or minor changes to the spec that can be ignored by older implementations can simply be filed as a github issue. + +## Proposal lifecycle + +Discuss -> Draft/POC -> Implementation -> Approval + +### Discuss +Start a [DISCUSS] thread on the mailing list (dev@parquet.apache.org) with your idea. At this point, the community can discuss whether the impact of the proposal requires a document here or just be a github issue. +Once you have a better idea of the general consensus on the proposal, open a github issue using the [proposal template](proposals/_PROPOSAL_TEMPLATE.md) +Attaching a google doc to collect feedback and collaborate with the community works usually well early on. + +*Transition:* Once you feel you received enough feedback or need to start the POC to have better answers to questions you get, you can move to the next step. Anybody is free to start POCs anytime. We just recommend getting feedback before you spend a significant amount of your time. + +### Draft/POC +Once you feel the discussion has stabilized and you are ready to start a POC, open a PR to add the proposal to the table in the [Active Proposals](#active-proposals) section bellow. Link all relevant discussion documents in the body of the corresponding github issue. If using google docs, make sure docs are owned by an account that will maintain them publicly (preffer personal account over a work account). Alternativaly you can create a new Markdown file in the proposals folder and give more visibility to the work in progress (see [the example](1_BASE64_ENCODING.md) ). +The proposal document can evolve along the course of the POC. In particular to add more links to findings and performance evaluations. Collaboration is encouraged. More validation on the POC increases the chances of success. + +Example: [https://github.com/apache/parquet-format/pull/221] + +Make sure you consider the [requirements document](https://docs.google.com/document/d/1qGDnOyoNyPvcN4FCRhbZGAvp0SfewlWo-WVsai5IKUo/edit?tab=t.0#heading=h.v4emiipkghrx) to ensure the success of the POC. (Note: this doc would become a markdown page in the repo) + +*Transition:* There is enough clarity on the spec for the new feature and we have identified the 2 initial reference implementations for verification. + +### Implementation +Once we have reached enough consensus on the formalized spec change and validated it through the POC, we should have a clear idea of whether we want to pursue the implementation across the ecosystem. +At this stage we should finalize a formal spec contribution to parquet-format and we need to meet the contribution guidelines to consider the implementation finished. +See [CONTRIBUTING guidelines](https://github.com/apache/parquet-format/blob/master/CONTRIBUTING.md#additionschanges-to-the-format). + +*Transition:* A PMC vote will formalize that we have concluded the implementation and are ready to release. + +### Approval +Once the implementation phase is finished, we can include the contribution in the next release. Congrats! + +## Active Proposals + +| ID | Description | Status | +|-----|--------------|---------| +| [github issue] | adding this new encoding | POC | +| [github issue] | add Variant type | Implementation | + +(Those are examples to be removed as we start using this) + +## Implemented +| ID | Description | Status | release it was added | +|-----|--------------|---------|-----------------------| +| [github issue] | encryption | Completed | x.y.z | + +(Those are examples to be removed as we start using this) + +## Archived + +| ID | Description | Status | reason for archiving | +|-----|--------------|---------|-----------------------| +| [github issue] | [adding base64 compression](1_BASE64_ENCODING.md) | Archived | POC showed that compression ratio was not practical | + +(Those are examples to be removed as we start using this) diff --git a/proposals/_PROPOSAL_TEMPLATE.md b/proposals/_PROPOSAL_TEMPLATE.md new file mode 100644 index 000000000..b55fa1011 --- /dev/null +++ b/proposals/_PROPOSAL_TEMPLATE.md @@ -0,0 +1,43 @@ + +# Proposal + +--- +Author: ~your name~ +Created: ~date~ +Name: *short sentence describing the proposal* +Issue: https://github.com/apache/parquet-format/issues/NNN +Status: DRAFT|IMPLEMENTATION|COMPLETED +--- + +## Description +*Short description of the proposal. Is it a new encoding? Is it backwards compatible (old readers will just ignore it)? Is it additional metadata?* + +## Rationale +Describe why this is a feature that will improve the parquet format and what alternatives currently exist for the use case (e.g. must use a different format, or "must build additional infrastructure to avoid re-parsing footer on each query", or "must use a general purpose compression algorithm to achieve the same space, thus slowing down query performance") + +## Spec + +At the proposal stage you don't need a fully fleshed out spec yet. +Please add any link to relevant documentation, papers, etc. +at the implementation stage, the details will need to be all clarified. + +## Evaluation +What datasets is it tested on and what is a success criteria +Please add any link to the relevant codebase.