Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Draggable objects

redblobgames.com

Post reply on HN
Print view View original

51–60 of 147 posts

Re: Draggable objects

#51

Post by dang »

sh1mmer wrote: post #50
dang wrote: post #47

Earlier quoted context omitted.

I've changed to that page from https://www.redblobgames.com/ . Thanks! Obviously, many comments on the thread predate this change, but there's enough context here for readers to figure that out, and hopefully we can get a more specific discussion going.

FWIW, dang, while I know guidelines are to post specific things for me personally “check out this collection of interactive tutorials” is actually a lot more interesting/helpful than linking to a single tutorial, especially when it’s not clear from that tutorial it’s part of a collection larger collection. Generally I feel like links to collections of stuff do well when they are interesting and don’t when they aren’t…

I hear you and am certainly not denying the usefulness of the site! it's fabulous and has been fabulous for many years. It has also made many great appearances on HN over the years: https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que....

But if we're to optimize HN for intellectual curiosity (https://hn.algolia.com/?dateRange=all&page=0&prefix=true&sor...), we have to consider thread quality, and there's no doubt that submissions like this generally lead to generic, and therefore shallow, discussion in the way that I described upthread.

Overall I think the best way for HN readers to discover a site like this is bottom-up: to run across an example of a great article and a great thread about it, and then click around to discover what else is there. This is more in the intended spirit of HN.

Edit: it's a little unorthodox for us to change the URL in midstream after a submission has this many upvotes and pre-existing comments, but I hope everyone understands that I did so to give the site more exposure and appreciation, not less. The alternative would have been to downweight the post as a "list submission" (https://news.ycombinator.com/item?id=37707904), and I didn't want to do that.

Re: Draggable objects

#52

Post by qiller »

Great write up for all the pitfalls and gotchas that come up when dealing with proper interactions

For something more "out of the box", I've been using interactjs for quite a while for a variety of my projects

Re: Draggable objects

#54

Post by trees101 »

Do graph search algorithms have applications for strategy? https://www.redblobgames.com/pathfinding/a-star/introduction... Consider a game like age of empires. The cost of attacking a village goes up when a wall is built around it. But when canon are invented, the cost goes down. Can we model each map (before and after invention of canon) and use it to plan a strategy? Same for building a port , portal or airport.

Re: Draggable objects

#55

Post by zem »

calebm wrote: post #30

Their hexagon grid page is one of the best hexagon grid resources on the web that I've found.

yeah! i clicked through the link and thought "oh, the hex grid guy!"

Re: Draggable objects

#56

Post by tsumnia »

gowld wrote: post #42
tsumnia wrote: post #28

A fantastic site. When I originally took over teaching Intro to AI, I initially relied on the A* search closed/open set pseudocode explanation[1]. However, when it would come time to ask students to implement it, I was constantly finding students absolutely confused by the approach. Once I swapped over to Amit's A* explanation, the number of confused students dropped significantly. Forever thankful for their walkthro…

What's the difference between the approaches?

Amit describes the differences here [2], but briefly:

- The Wiki pseduocode uses Set data structures for everything. While these are covered in classes, the up-tree concept isn't as heavily described as other trees

- Looking at it now, it looks like the pseudocode is a little more beginner friendly, but "back in mah day" it was not

- Although, RedBlob's pseudocode for obtaining neighboring nodes is clearer than "for each neighbor of current" (Wikipedia pseudocode)

- RedBlob uses Priority Queues and Maps instead of Sets, which connect better to other AI searches / recommendation algorithms. Higher 'priority' recommendations move to root in PQs while nodes in Sets... don't... they just sort of 'exist'

- The Maps make looking up node costs more intuitive than lists (aka "current := the node in openSet having the lowest fScore[] value")

[2] https://www.redblobgames.com/pathfinding/a-star/implementati...

Re: Draggable objects

#57

Post by VHRanger »

trees101 wrote: post #54

Do graph search algorithms have applications for strategy? https://www.redblobgames.com/pathfinding/a-star/introduction... Consider a game like age of empires. The cost of attacking a village goes up when a wall is built around it. But when canon are invented, the cost goes down. Can we model each map (before and after invention of canon) and use it to plan a strategy? Same for building a port , portal or airport.

That's effectively search on the decision tree. There's a lot of research on this!

Look into Montecarlo tree search, CFRM, AB pruning, and also more recent deep learning methods.

It's a very exciting area of research

Re: Draggable objects

#58

Post by amitp »

gowld wrote: post #42
tsumnia wrote: post #28

A fantastic site. When I originally took over teaching Intro to AI, I initially relied on the A* search closed/open set pseudocode explanation[1]. However, when it would come time to ask students to implement it, I was constantly finding students absolutely confused by the approach. Once I swapped over to Amit's A* explanation, the number of confused students dropped significantly. Forever thankful for their walkthro…

What's the difference between the approaches?

1. I use words for variable names instead of single letters like the textbooks do. I use "priority" instead of "F", "cost_so_far" instead of "G", "heuristic" instead of "H", "cost" instead of "w", "frontier" instead of "OPEN" or "O", "visited" instead of "CLOSED" or "C", "current" instead of "u", "next" or "neighbor" instead of "v".

2. The textbooks use an "open" and "closed" set. But in code, these aren't explicitly stored in set data structures. Instead, they're implicit. The cost_so_far dict(map) contains as keys both the open and closed sets, and the frontier (priority queue) contains the open set. So in my explanation of A* I focus on these data structures (priority queue and dict) instead of the open/closed sets. And when I do talk about the sets, I talk about the combined open and closed sets, calling it "visited" or "reached", because it's the combined set that is actually in the data structures.

3. The textbooks use a priority queue with reprioritization. When you visit a node that has a lower cost than the previously found cost, you go into the priority queue and adjust the cost. In my presentation I don't use reprioritization. Instead, I insert another entry into the priority queue with the lower cost. This makes the priority queue simpler (reprioritization is complicated). And in practice, I think it's faster too.

Re: Draggable objects

#59

Post by amitp »

tsumnia wrote: post #28

A fantastic site. When I originally took over teaching Intro to AI, I initially relied on the A* search closed/open set pseudocode explanation[1]. However, when it would come time to ask students to implement it, I was constantly finding students absolutely confused by the approach. Once I swapped over to Amit's A* explanation, the number of confused students dropped significantly. Forever thankful for their walkthro…

That's great to hear — thank you!
Post reply on HN