Sorawee Porncharoenwase
@sorawee_p
Followers
228
Following
4K
Media
95
Statuses
1K
Joined June 2019
RT @ellisk_kellis: New paper: World models + Program synthesis by @topwasu.1. World modeling on-the-fly by synthesizing programs w/ 4000+ l….
0
105
0
Generate videos in just a few seconds. Try Grok Imagine, free for a limited time.
384
672
3K
Better garbage collection in NYC? Is this sponsored by Jane Street?.
2
1
29
No?.
Learning how to program quantum computers is the next critical skill for developers. Learn more about the state-of-the-art and challenges in this area at the Workshop on Quantum Software 2024 on Monday, June 24. Full program here:
0
0
2
Collatz conjecture is joked as a Soviet conspiracy to slow down American mathematics. I recall there's another {physics,math,optimization} problem that is the inverse: American scientists suggested exporting the problem to waste time of the enemy's scientists. What's the problem?.
0
1
1
Is it just me? People asked for programming/math help, saying their solution is wrong. When I ask for their understanding, they have none, and it turns out their wrong solution is ChatGPT-generated. Suddenly I no longer want to help them.
4
0
17
Art project idea: an interior design (or anything really) that makes you feel, this is generated by GenAI.
0
0
1
RT @eeide: The video of my presentation at GPCE '23 is now online. Learn about generating "conforming" programs with Xsmith! https://t.co/….
0
7
0
Artifacts that need a lot of resources should use SNARK.
0
0
0
But yeah, let me know if you have any feedback. Also, thanks to @d_christiansen and @yminsky for the suggestions that made me implement the feature!. two_columns documentation: Original paper: 18/18.
dl.acm.org
Pretty printers make trade-offs between the expressiveness of their pretty printing language, the optimality objective that they minimize when choosing between different ways to lay out a document,...
0
0
1
I had a lot of failed attempts (e.g. optimality is subtly violated), because conditions that are previously true no longer hold when two_columns is added. 17/.
1
0
0
Implementation-wise, this is the most complicated feature in PrettyExpressive so far, doubling the size of the pretty printer. It is also very tricky to get it right. 16/.
1
0
0
And since two_columns doesn't create a table layout, formatting stuff like 2D syntax ( without hard coding content is impossible for the moment. 15/
1
0
0
Also, it's only 2 columns. fill and fillBreak OTOH can be chained (assuming no overflow) to create >= 3 columns. But given how expensive two_columns already is, I doubt that e.g. three_columns will work well. I think two_columns will already suffice for most uses however. 14/.
1
0
0
I actually have an idea for how to improve this, though it will still be expensive regardless. /13.
1
0
0
What's the catch? This feature is very expensive! Each two_columns with N rows *at the very least* expands to a document of size N^2. But that's what you pay for uncompromised optimality and more expressiveness. It should work practically well for a small # of rows tho. 12/.
1
0
0
Good news: PrettyExpressive allows you to tell what you care via its cost factory! Simply make the penalty for exceeding the col sep higher than the penalty for exceeding the page limit, and now it will pick your desired layout! There are more knobs that you can adjust too. 11/
1
0
0
As shown previously, PrettyExpressive will first and foremost avoid overflow over the page limit. But what if you don't care about that as much, compared to getting the right column aligned properly at the column separator? 10/.
1
0
0