diff options
Diffstat (limited to '2021/talks/bindat.md')
| -rw-r--r-- | 2021/talks/bindat.md | 83 |
1 files changed, 83 insertions, 0 deletions
diff --git a/2021/talks/bindat.md b/2021/talks/bindat.md new file mode 100644 index 00000000..e9cfffaa --- /dev/null +++ b/2021/talks/bindat.md @@ -0,0 +1,83 @@ +[[!meta title="Turbo Bindat"]] +[[!meta copyright="Copyright © 2021 Stefan Monnier"]] +[[!inline pages="internal(2021/info/bindat-nav)" raw="yes"]] + +<!-- You can manually edit this file to update the abstract, add links, etc. ---> + + +# Turbo Bindat +Stefan Monnier + +[[!inline pages="internal(2021/info/bindat-schedule)" raw="yes"]] + + +# Table of Contents + + + +Bindat is an ELisp library to help manipulate binary data. This is a +niche library that is used by packages such as Websocket, EMMS, and +cpio-mode. Its implementation was repeatedly caught harassing hapless +kitten while at the same time providing poor service slowly. For +Emacs-28, Bindat was rewritten so as to make it more efficient and +flexible while respecting the kitten. In this presentation I intent to +show how we saved those. Not recommended for birds. + +# Discussion + +- Q1: bindat seems very similar to GNU Poke (except that GNU Poke is a + superset, and then some, with a different syntax). I'm wondering if + it might be good to add a bindat variant that translates to/from + Poke if need be (using libpoke), for sheer insane blazing + native-code JITted speed. (And, later, maybe letting bindat gain + some of the insanely expressive capabilities GNU Poke has got). Its + use of eval blocked this in times past. but now... + - A:GNU Poke is indeed the natural evolution, and is much more + powerful. Given the fairly little use of BinDat so far, I'm + not sure there will be enough motivation to give access to GNU + Poke from Emacs, tho. One of the main benefits of using GNU + Poke would probably be that lots of formats are already + available for GNU Poke, so you could directly re-use them. +- Q2: Is your dog's name something Lisp or PL related...? :) + - A:Winnie? I don't think so, no (we didn't choose the name, in + any case) +- Q3: This looks amazing! Is it merged into mainline Emacs, a patch, + an external library? + - A: It's in Emacs-28 +- Q4: Are there benchmarks of this vs. the older bindat? + - A:There is a benchmark for it in the `elisp-benchmarks` +- Q5: Do you know of any CL or Scheme libs similar to bindat.el? + - A: No, but I'd be interested to hear about it if someone else + does. +- Q7: You are a hero of kittens everywhere. Do you have any feline + pets as well? :) + - A: Not yet. If you're near Montreal and you have a kitten for + me, I'm interested +- I *hope* cl-loop is more efficient than building a bunch of intermediate lists when you chain map/filter/reduce operations. +- Curious: how is gnu poke more flexible? +- What hobbies/interests do you have besides Emacs (and PL)? :) +- do you have any thoughts about how to make EmacsConf even better next year? +- I was surprised to see that a whole new DSL was developed for poke from scratch. Do you think would have been better to develop/improve a library like bindat on top of an existing language instead? +- What are some of your favorite talks from this conf so far? +- what kind of dog is Winnie? + - comment: I hadn't heard of that breed before +- How do you see more control over types (type hints/decl through type specifiers etc) (SBCL like programming model) coming into Elisp? +- Do you plan to add bit-level support? + +Other comments: + +- I can imagine using bindat to improve Emacs's music player packages +- yes last year the Q&A periods were much longer + - last year some of the presentations were live though +- I've asked this question to them during LPC 2020 but infact haven't got a very satisfactory answer :) +- If you ever write a library for window management in Emacs, you could call it winnie.el :) +- hints in unoptimized code should be assertions +- we probably need both ways of compiling: safe and less safe :) +- I think this is classic problem that is almost impossible to accomplish. many libraries try to do that but in the end the only working ones are relaying on C compilers. +- also you have the problem of size of objects. like how big is a long? this is not specified and is arch dependent +- parsing a generic .h file is way more difficult but is another subject. +- yep, the automatic translation is more for libraries trying to write automatically C bindings + +[[!inline pages="internal(2021/captions/bindat)" raw="yes"]] + +[[!inline pages="internal(2021/info/bindat-nav)" raw="yes"]] |