mirror of
https://github.com/github/semantic.git
synced 2024-12-18 20:31:55 +03:00
65 lines
5.6 KiB
Markdown
65 lines
5.6 KiB
Markdown
#09-30-2018
|
|
|
|
**What went well?**
|
|
|
|
@tclem: Quiet week with everyone in Japan. Was heads down in parsing issues, identifying determiners. Tackled some other harder issues (ongoing now around tooling and command line parsing).
|
|
|
|
@joshvera: ICFP was spectacular. There's a lot to write down, and ideas that could potentially apply to future features or techniques in general that would be useful. Testing RWS has gone well. Issue came up with small RWS / SES diff ordering.
|
|
|
|
@rewinfrey: ICFP also amazing. This week small issues and started looking at keeping parent contexts for nested contexts (and have a RFC out).
|
|
|
|
@pengwynn: Good to interface with everyone, and to see if we provide further support with platform and integrating with dotcom.
|
|
|
|
**What were the challenges?**
|
|
|
|
@tclem: Challenging being totally on my own a couple weeks into the project. Some issues came up that required some heavy investigation (like monad transformers).
|
|
|
|
@joshvera: Coming back has been surprisingly difficult to have normal hours (going to Japan was easy). RWS bug with ordering terms is concerning, but should be manageable. Some PRs on master fail and think they are problems that come up in our property tests rarely (should look back and get those seeds so we can reproduce them). We have to make our diffs commute, so that if we're producing the same diff terms, they should always be in the same order. Sometimes surrounding context will change or alter diffs.
|
|
|
|
@rewinfrey: I'm not looking at RWS so things are going well :p
|
|
|
|
**What did you learn?**
|
|
|
|
@tclem: Studying folds and algebraic data types. Learning about binary tree operations. Understanding of the entire architecture stack.
|
|
|
|
@joshvera: Learned a lot at ICFP - specifically about effects and coeffects (reifying as types in a language). Effects do something to a context, coeffects reach into and look at a context. For example, a high security program that communicates with a low security program could allow the low security program to use a coeffect to understand something about what happened in the high security program, but wouldn't reveal the sensitive information.
|
|
|
|
@rewinfrey: Learned a lot about freer monad and effects in general. Another big take away was homotopy type theory.
|
|
|
|
----
|
|
|
|
@pengwynn: So I'd like to ask what are we missing on the team?
|
|
|
|
@joshvera: The help we get from dotcom (mclark and brianmario) is very helpful so we don't have to spend as much time diving into dotcom. We haven't received a lot of design attention - have a couple technical reasons why semantic diff is currently disabled. Would like to follow up with Fabian on design next week. Would like to find people with more Haskell experience, or with stronger background in diffing in general outside of the context of version control. One of the bigger technical areas of work we will have to do is writing parsers and ensuring they are correct.
|
|
|
|
@pengwynn: How do we leverage data to make autonomous coding a reality? As we move closer to 2017 what's the team make up look like?
|
|
|
|
@joshvera: We're getting to a point where semantic diffing is stable. Figuring out a way to generate a corpus from the information we have would be a really hard technical solution, but would pay back in dividends. Things like semantic merging. Also Rob has ideas about how we would compare structures across different languages, and code navigation (semantic click through to definition). What are the most useful features for people we could develop immediately? It would be good to have a stronger interface with product people than we currently do.
|
|
|
|
@tclem: We could have a dedicated effort on parsing.
|
|
|
|
@pengwynn: Is that a polyglot or Haskeller?
|
|
|
|
@joshvera: That's someone like Max (Max Brunsfeld) at the moment -- but in general that means a polyglot.
|
|
|
|
@tclem: I would +1 for fabian or product design. Having someone that can help us understand the context of the future of PR's, and where that general set of features are going.
|
|
|
|
@pengwynn: What would it take once we launch diff summaries to go from 2 languages to 25? If that's one hire we can make that happen. Want to
|
|
poll the group to determine what our holes are and what we should look for when hiring for the next teammate.
|
|
|
|
@tclem: Looking at parsers is a good thing.
|
|
|
|
@joshvera: Getting someone to be able to write parsers from a specification (which is researched). Taking existing stuff and putting it into a system that works for us. Also integrating with existing compilers, so we don't have to manually update our parsers every time a language change would be very beneficial. Automating that problem away would be beneficial, too. Hooking into different compilers would be something that would save us effort and energy in the future.
|
|
|
|
@pengwynn: This is awesome, and spurs some conversations with JD's team.
|
|
|
|
----
|
|
|
|
@rewinfrey: Where are we with getting diff summaries staff shipped and also out to gen pop? Would it be good to have a conversation about that early next week to clarify the direction for prioritization?
|
|
|
|
@joshvera: RWS and platform interface issues are the two main things preventing us from staff shipping again. Also want to talk with product design to take existing feedback and adjust accordingly. There are a few things in the summer eyes milestone that I think would be most useful, but maybe those aren't what product thinks is most useful. There could be more tiny bugs that only come up in certain contexts that we might not be aware of since we only do 100 property tests.
|
|
|
|
@tclem: It feels like we're close, and that it's really product and design and incorporating diff summaries into the new focus on code review.
|
|
|
|
@joshvera: I'm planning to talk with Fabian early next week to sync up on design.
|