SIP Meeting Minutes - 13th July 2016

Language

Minutes

The following agenda was distributed to attendees:

Topic Reviewer
Discussion of the new SIP process Jorge Vicente Cantero
SIP 25 - Trait parameters Adriaan Moors
SIP 26 - Unsigned Integer Data Types Martin Odersky
SIP 22 - Async Eugene Burmako
SIP 20 - Improved lazy val initialization Sébastien Doeraene
Trailing commas SIP Eugene Burmako

Quick iteration through all the SLIPs:

Jorge Vicente Cantero was the Process Lead and acting secretary of the meeting.

The following proposals were numbered:

  • SIP-26: Unsigned Integer Data Types
  • SIP-27: Trailing commas

(When a SIP is numbered, it can be thought of as a first-round of acceptance. That is, the committee has voted in favor of the changed being accepted into Scala in theory, so long as all potential design and implementation flaws are eventually addressed and worked through. Typically, the committee will raise a number of important concerns about the SIP that must be addressed, as next steps, ideally before the next meeting of the SIP committee.)

The following other proposals were discussed:

  • SIP-22: Async (postponed)
  • SIP-20: Improved lazy val initialization
  • SIP-25: Trait Parameters

Some other library proposals were evaluated and the committee gave feedback to the authors.

Date, Time and Location

The meeting took place at 5:00pm Central European Time / 8:00am Pacific Daylight Time on Wednesday, July 13th, 2016 via Google Hangouts.

Minutes were taken by Jorge Vicente Cantero, acting secretary.

Attendees

Attendees Present:

Guests

Proceedings

Opening Remarks

As acting Process Lead, Jorge Vicente Cantero conducted the meeting, made the opening remarks, and introduced the guest Dmitry, who was present to help discuss the proposal for an improved lazy val initialization (SIP-20).

Scala Improvement Proposals

Proposal SIP-25: Trait Parameters proposed by Martin Odersky

Adriaan Moors, as the assigned reviewer of this SIP, quickly introduced the proposal. The proposal helps to abstract over traits by introducing type parameters, a feature that was only possible in classes.

Adriaan points out that it needs a little bit more of work. He generally advises to give more details about how the proposed changes interact with other features. In concrete, he’d like to know what the modifiers mean, and what would happen if there’s an implicit modifier. On a side note, he thinks that there should be some guidelines on how the proposal impacts programmers and what technical issues are addressed. He thinks that it would be great to see an implementation, as the one in Dotty. He considers this proposal is a good candidate for 2.13.

Martin and Heather also discuss what the role of a reviewer is. Jorge clarifies that technical discussions should take place in the meeting.

Outcome: The board agreed to schedule the next iteration of the evaluation process in 6 months, since there’s no implementation yet and the authors need time to produce one.

Proposal SIP-26: Unsigned Integer Data Types by Denys Shabalin and Sébastien Doeraene

Martin is the reviewer of this SIP. He’s on the fence of accepting this proposal, he would prefer to see it in the platform as a library, since putting it in the core would require too much work and he’s unsure if that would be a priority.

Sébastien, one of the authors, points out that placing it as a library defeats the purpose of the SIP (because cooperative equality would not exist), which is to allow the native platforms to benefit from it (Scala.js and Scala Native). He explains that, in order to make it a library, he would need at least two SIPs to make it interact correctly with Scala.js (the value classes formalization is not suitable for what he wants to address).

Dmitry, Martin and Sébastien start to discuss about the performance of other alternatives that would need to change the representation of scala number. Adriaan and Josh agree that the proposal would be better as a library.

Outcome: The board voted; all were in favor of giving it a number. Jorge asks the authors to make a PR to the SIP website repo. The next iteration would be in September because Sébastien is on vacation in August. He needs to prepare its evaluation in September by tweaking the changes in BoxesRunTime.scala so that the performance of existing codebases does not suffer any degradation, and so that the performance of non-unsigned integer comparisons is not affected (or very little) by the unrelated addition of unsigned integers in the codebase.

Proposal SIP-22: Async proposed by Philipp Haller and Jason Zaugg

Eugene Burmako does a thorough description of the SIP and describes its historical background. He roughly talks about the implementation, which uses macros, and he’s impressed of its quality in the design and implementation.

Other languages like F#, C# and JS have something similar. There’s a restriction that the functionality cannot be used inside a try catch. Eugene reveals that the authors have asked for a timeout to improve the implementation and the design. He recommends them to add more documentation as in C# and suggests to close it and wait until the authors resubmit it.

Jorge and Heather discuss about what are the differences between postponing and marking a SIP as dormant. The idea is that SIPs marked as dormant are the ones that have been evaluated, but there hasn’t been any activity in two months. Postponing a SIP is done when we know beforehand that some constraints need to be resolved before resuming its evaluation.

Outcome: The Process Lead postpones it until the authors want to decide to revisit the support of async/await in try/catch blocks. When that’s considered, this SIP should be reopened and it should see another round of discussion.

Proposal SIP-20: Improved lazy val initialization presented

Sébastien reviews the SIP and asks Dmitry, present in the meeting, to correct him if he’s wrong. He agrees that the SIP is desirable but he’s unsure about the benchmarks and which of the proposals is faster. Dmitry explains that the benchmarks are in the repository. Sébastien also points out that there’s an implementation missing for scalac, and recommends the author to include more documentation..

Outcome: For the next iteration, the reviewer suggests that the SIP should have an updated specification, implementation and benchmarks. The Process Lead schedules the next iteration by October 2016.

Proposal SIP-27: Trailing Commas

Eugene Burmako, who also reviewed this recently submitted SIP, explains what the proposal addresses. The proposal seeks to introduce changes in the syntax of the language that will not error when commas are placed in concrete valid places. He makes the point that it has several benefits; for instance, diffs in github will only show one changed line when a new element is added in a list whose elements are placed in independent lines.

He also discusses that there are some issues with the interaction of Tuple1 and pretty printing. The proposal is minor but addresses day-to-day annoyances. Martin fears that this proposal would interfere with another important future SIP that will integrate generic programming with Scala. Adriaan doesn’t like the idea. Josh proposes to unify tuples with other features of the language, like parameters lists and the apply methods.

Adriaan wants to wait for the proposal of how to do generics over tuples, and integrating hlists with Scala, which he thinks it’s the really important proposal.

Outcome: 2 people abstain, 3 people vote in favor of it. Josh’s connectivity drops out, and he’s not able to vote. The Process Lead decides to give it a number. Authors are asked to prepare for the first iteration of the evaluation process in August. This involves exploring interactions with other language features by exhaustively enumerating the locations in the grammar where trailing commas may be used.

SLIPs

Jorge asks the SIP committee to provide feedback to the authors to speed up the SLIP process in the future.

  • JSON AST: No news from the last discussion in the slip repo. It’s been integrated into Play and sbt server 1.0. The committee considers that it’s a prime candidate for the platform.
  • Extensions of Futures and Promises: the committee calls for an implementation. This SLIP will be addressed by the next SLIP committee.
  • Either monadic bias: it was merged one day before by the Lightbend team. Therefore, this SIP is both accepted and merged.
  • scala.io.Target: Martin proposes to delay it until the SLIP committee decides how the platform would look like. Then, they will take care of it.
  • Redesigning collections views: Jorge explains what Josh, who is author of the SLIP, proposes for its design:
    • Iterator-based API (supports join and efficient deferred operation with one time parse)
    • Transducer-based API (supports non-iterable collections and efficient deferred operations, but doesn’t support co-iteration, i.e. efficient join). Martin and Dmitry discuss that iterators are more powerful, and are looking forward to an implementation. Therefore, authors are asked to provide one for the first iteration in the SLIP meeting.

Other business

Jorge confirms that there will be a new SLIP process proposed for the middle of August.

Closing remarks

See you next time!