A common bug-inducing pattern: building a grid out of linked nodes.
-
harold aptroot wrote:
Perhaps the funniest part is that while all that is going on, it is realised that a 2D array is needed in order to quickly find a node with given coordinates, and from that point on all those buggy links are worse than useless: the location in the array implicitly defines the neighbours.
Arrays are the worst choice of all. How about a quadtree for 2D structures or octrees for 3D structures?
I have lived with several Zen masters - all of them were cats. His last invention was an evil Lasagna. It didn't kill anyone, and it actually tasted pretty good.
-
harold aptroot wrote:
I never said what you imagined I said. The "removing" thing is in the context of nodes.
If you implement the map as a list of nodes, you'd still use a flag instead of removing the nodes.
harold aptroot wrote:
Anyway I think you can reasonably assume that I'm not a total idiot.
I don't like assumptions, and don't know how your coworkers think. I can also state that I am a certified idiot; that's why I need specifications, and come asking questions that may seem obvious to others. Let them play Rimworld, might help :)
Bastard Programmer from Hell :suss: If you can't read my code, try converting it here[^] "If you just follow the bacon Eddy, wherever it leads you, then you won't have to think about politics." -- Some Bell.
-
OK sure, there would be a flag. Do I really need to specify implementation details like that though?
harold aptroot wrote:
OK sure, there would be a flag. Do I really need to specify implementation details like that though?
Not required, it just helps me understand what the problem is. Might be beneficial if I encounter it myself. My apologies if I'm not helping in any way :thumbsup: If you have a flag and don't remove nodes, how can there be invalid (double/missing) nodes?
Bastard Programmer from Hell :suss: If you can't read my code, try converting it here[^] "If you just follow the bacon Eddy, wherever it leads you, then you won't have to think about politics." -- Some Bell.
-
No way, arrays are great. Quadtrees are fun and powerful, but at the cost of being at least an order of magnitude more complicated.
harold aptroot wrote:
being at least an order of magnitude more complicated
Which ltttle object oriented me would just hide by properly encapsulating it away in methods and properties.
I have lived with several Zen masters - all of them were cats. His last invention was an evil Lasagna. It didn't kill anyone, and it actually tasted pretty good.
-
harold aptroot wrote:
being at least an order of magnitude more complicated
Which ltttle object oriented me would just hide by properly encapsulating it away in methods and properties.
I have lived with several Zen masters - all of them were cats. His last invention was an evil Lasagna. It didn't kill anyone, and it actually tasted pretty good.
-
harold aptroot wrote:
OK sure, there would be a flag. Do I really need to specify implementation details like that though?
Not required, it just helps me understand what the problem is. Might be beneficial if I encounter it myself. My apologies if I'm not helping in any way :thumbsup: If you have a flag and don't remove nodes, how can there be invalid (double/missing) nodes?
Bastard Programmer from Hell :suss: If you can't read my code, try converting it here[^] "If you just follow the bacon Eddy, wherever it leads you, then you won't have to think about politics." -- Some Bell.
-
Exactly, that's the point: there would be no possibility of creating invalid states that way. Any state that can be represented is valid .. it might be the wrong state, but it cannot be internally inconsistent.
harold aptroot wrote:
Exactly, that's the point: there would be no possibility of creating invalid states that way. Any state that can be represented is valid .. it might be the wrong state, but it cannot be internally inconsistent.
That's how most games will approach it. Sometimes there's water or land in the Rimworld map where there shouldn't be - but the map in itself is "valid" and hence, playable.
Bastard Programmer from Hell :suss: If you can't read my code, try converting it here[^] "If you just follow the bacon Eddy, wherever it leads you, then you won't have to think about politics." -- Some Bell.
-
Hiding complexity under the rug doesn't make it go away, you'd still have to actually implement that quad tree.
Sure, but I have already done that a few times and I never had much trouble managing a few pointers and memory. Anyway, in many languages overstepping the boundaries of an array may have worse consequences than just causing an exception. Your implementation therefore also offers the advantage to allow as little or as much error handling as is needed.
I have lived with several Zen masters - all of them were cats. His last invention was an evil Lasagna. It didn't kill anyone, and it actually tasted pretty good.
-
Sure, but I have already done that a few times and I never had much trouble managing a few pointers and memory. Anyway, in many languages overstepping the boundaries of an array may have worse consequences than just causing an exception. Your implementation therefore also offers the advantage to allow as little or as much error handling as is needed.
I have lived with several Zen masters - all of them were cats. His last invention was an evil Lasagna. It didn't kill anyone, and it actually tasted pretty good.
CodeWraith wrote:
Anyway, in many languages overstepping the boundaries of an array may have worse consequences than just causing an exception.
Not too hard to write a method that takes any value and checks for overflows. If you overflow on the right of the map, set the value to zero, and you re-appear on the left of the map.
Bastard Programmer from Hell :suss: If you can't read my code, try converting it here[^] "If you just follow the bacon Eddy, wherever it leads you, then you won't have to think about politics." -- Some Bell.
-
CodeWraith wrote:
Anyway, in many languages overstepping the boundaries of an array may have worse consequences than just causing an exception.
Not too hard to write a method that takes any value and checks for overflows. If you overflow on the right of the map, set the value to zero, and you re-appear on the left of the map.
Bastard Programmer from Hell :suss: If you can't read my code, try converting it here[^] "If you just follow the bacon Eddy, wherever it leads you, then you won't have to think about politics." -- Some Bell.
Map? Reappear on the other side? How primitive! I have a map of a galaxy with 4 billion x 4 billion x 4 billion coordinate points and who knows how many hundreds of thousands of stars. There are 4 billion of these galaxies in each of the 4 billion universes. One of the galaxies should last the players forever. Don't try to store all that in an array. :-)
I have lived with several Zen masters - all of them were cats. His last invention was an evil Lasagna. It didn't kill anyone, and it actually tasted pretty good.
-
Map? Reappear on the other side? How primitive! I have a map of a galaxy with 4 billion x 4 billion x 4 billion coordinate points and who knows how many hundreds of thousands of stars. There are 4 billion of these galaxies in each of the 4 billion universes. One of the galaxies should last the players forever. Don't try to store all that in an array. :-)
I have lived with several Zen masters - all of them were cats. His last invention was an evil Lasagna. It didn't kill anyone, and it actually tasted pretty good.
CodeWraith wrote:
Map? Reappear on the other side? How primitive!
May be, but it is simple and it works :)
CodeWraith wrote:
I have a map of a galaxy with 4 billion x 4 billion x 4 billion coordinate points and who knows how many hundreds of thousands of stars. There are 4 billion of these galaxies in each of the 4 billion universes. One of the galaxies should last the players forever. Don't try to store all that in an array. :)
Tell me you are building Elite III or a new Master of Orion :D
Bastard Programmer from Hell :suss: If you can't read my code, try converting it here[^] "If you just follow the bacon Eddy, wherever it leads you, then you won't have to think about politics." -- Some Bell.
-
CodeWraith wrote:
Map? Reappear on the other side? How primitive!
May be, but it is simple and it works :)
CodeWraith wrote:
I have a map of a galaxy with 4 billion x 4 billion x 4 billion coordinate points and who knows how many hundreds of thousands of stars. There are 4 billion of these galaxies in each of the 4 billion universes. One of the galaxies should last the players forever. Don't try to store all that in an array. :)
Tell me you are building Elite III or a new Master of Orion :D
Bastard Programmer from Hell :suss: If you can't read my code, try converting it here[^] "If you just follow the bacon Eddy, wherever it leads you, then you won't have to think about politics." -- Some Bell.
I still play Master of Orion II very often. I have not really worked on my game in quite a while now. It's more a browser game that has accidentally fallen out of the browser. I have posted the link to the video before, but you can see what I need octrees for: link[^]
I have lived with several Zen masters - all of them were cats. His last invention was an evil Lasagna. It didn't kill anyone, and it actually tasted pretty good.
-
I still play Master of Orion II very often. I have not really worked on my game in quite a while now. It's more a browser game that has accidentally fallen out of the browser. I have posted the link to the video before, but you can see what I need octrees for: link[^]
I have lived with several Zen masters - all of them were cats. His last invention was an evil Lasagna. It didn't kill anyone, and it actually tasted pretty good.
Have you considered a pre-release on Steam? If you need someone to test the game, please let me know :D
Bastard Programmer from Hell :suss: If you can't read my code, try converting it here[^] "If you just follow the bacon Eddy, wherever it leads you, then you won't have to think about politics." -- Some Bell.
-
Have you considered a pre-release on Steam? If you need someone to test the game, please let me know :D
Bastard Programmer from Hell :suss: If you can't read my code, try converting it here[^] "If you just follow the bacon Eddy, wherever it leads you, then you won't have to think about politics." -- Some Bell.
I would first need to get my Padawan onboard again and then we need to set up a new server.
I have lived with several Zen masters - all of them were cats. His last invention was an evil Lasagna. It didn't kill anyone, and it actually tasted pretty good.
-
harold aptroot wrote:
Exactly, that's the point: there would be no possibility of creating invalid states that way. Any state that can be represented is valid .. it might be the wrong state, but it cannot be internally inconsistent.
That's how most games will approach it. Sometimes there's water or land in the Rimworld map where there shouldn't be - but the map in itself is "valid" and hence, playable.
Bastard Programmer from Hell :suss: If you can't read my code, try converting it here[^] "If you just follow the bacon Eddy, wherever it leads you, then you won't have to think about politics." -- Some Bell.
Just a common question - once you fit something as a grid/array, the mechanism to reach to connected notes is provided by the indices. When we have this, do we still need a "link" in nodes again to answer the same need? Once a grid structure comes in, I feel nodes can be completely agnostic about their neighborhoods, removing the in-built links? (I'm more stupid than you, so more questions) :)
Starting to think people post kid pics in their profiles because that was the last time they were cute - Jeremy Falcon.
-
Just a common question - once you fit something as a grid/array, the mechanism to reach to connected notes is provided by the indices. When we have this, do we still need a "link" in nodes again to answer the same need? Once a grid structure comes in, I feel nodes can be completely agnostic about their neighborhoods, removing the in-built links? (I'm more stupid than you, so more questions) :)
Starting to think people post kid pics in their profiles because that was the last time they were cute - Jeremy Falcon.
Vunic wrote:
When we have this, do we still need a "link" in nodes again to answer the same need?
Sounds redundant :)
Vunic wrote:
Once a grid structure comes in, I feel nodes can be completely agnostic about their neighborhoods, removing the in-built links?
Sounds cleaner, doesn't it?
Vunic wrote:
(I'm more stupid than you, so more questions) :)
So this is not a trick-question?
Bastard Programmer from Hell :suss: If you can't read my code, try converting it here[^] "If you just follow the bacon Eddy, wherever it leads you, then you won't have to think about politics." -- Some Bell.
-
By "grid" I just mean a good old 2D arrangement of cells like a chess board. Building it out of linked nodes means making every cell a node with explicit references to neighbouring node objects. This is something that I've seen novices do quite a lot more than necessary. There can be reasons to do it, but usually it results in an over-explicit and over-redundant data structure and plenty of bugs. Every degree of redundancy is an opportunity for the data structure to get out of sync with itself, and anything explicit can be explicitly incorrect. Usually the reason they give for doing it is "because path finding algorithms work on graphs", but graphs are an abstract interface, not an implementation. What tends to happen (obviously it is also possible to "just write correct code", but the point is it's easy to get wrong), especially if nodes are dynamically added and removed, is that the links decay over time. Nodes start disagreeing about whether they are connected, resulting in one-way connections, and the bad state starts to infect everything. Nodes that should have been disconnected entirely turn out to be linked to by some random other node. Multiple nodes that represent the same coordinate come into being. "Wild links" are teleporting all over the map for no clear reason. The path finding gets very confused and the programmer does too since what they see in the debugger (if they look at all, which many novices avoid for some reason) doesn't make any sense. Perhaps the funniest part is that while all that is going on, it is realised that a 2D array is needed in order to quickly find a node with given coordinates, and from that point on all those buggy links are worse than useless: the location in the array implicitly defines the neighbours. All problems could have been avoided by just working with that 2D array to begin with. The only way you have a node is because you had its coordinates, so you always know where its neighbours are without even storing the coordinates on the node object. There are no more explicit neighbour links, so they cannot be incorrect. Marking a cell in the grid as non-walkable implicitly deletes whatever links that needs to delete, it cannot go wrong since there are no actual links to actually delete. This phenomenon still surprises me. Building a grid out of linked nodes is something I had never even considered doing before I saw someone else making that mistake. Surely a 2D array is the first thing you think of when your data has the shape of a 2D array, that just makes sen
harold aptroot wrote:
A common bug-inducing pattern...This is something that I've seen novices do quite a lot more than necessar
This is "common" and done a "lot" where exactly? What industry? I have worked in a few and cannot ever recall that pattern even being in use. Not to mention that your solution is fixed size. How can someone add and remove nodes to a fixed size structure and yet manage to not correctly hook those up? And yet is something that only shows up over time? If they are directly manipulating the structure all over the place and not using a well defined API then certainly that would be the case but that would be true for any data structure entity, like lists, hashes even databases. Is that the real problem that there is no well defined API?
-
By "grid" I just mean a good old 2D arrangement of cells like a chess board. Building it out of linked nodes means making every cell a node with explicit references to neighbouring node objects. This is something that I've seen novices do quite a lot more than necessary. There can be reasons to do it, but usually it results in an over-explicit and over-redundant data structure and plenty of bugs. Every degree of redundancy is an opportunity for the data structure to get out of sync with itself, and anything explicit can be explicitly incorrect. Usually the reason they give for doing it is "because path finding algorithms work on graphs", but graphs are an abstract interface, not an implementation. What tends to happen (obviously it is also possible to "just write correct code", but the point is it's easy to get wrong), especially if nodes are dynamically added and removed, is that the links decay over time. Nodes start disagreeing about whether they are connected, resulting in one-way connections, and the bad state starts to infect everything. Nodes that should have been disconnected entirely turn out to be linked to by some random other node. Multiple nodes that represent the same coordinate come into being. "Wild links" are teleporting all over the map for no clear reason. The path finding gets very confused and the programmer does too since what they see in the debugger (if they look at all, which many novices avoid for some reason) doesn't make any sense. Perhaps the funniest part is that while all that is going on, it is realised that a 2D array is needed in order to quickly find a node with given coordinates, and from that point on all those buggy links are worse than useless: the location in the array implicitly defines the neighbours. All problems could have been avoided by just working with that 2D array to begin with. The only way you have a node is because you had its coordinates, so you always know where its neighbours are without even storing the coordinates on the node object. There are no more explicit neighbour links, so they cannot be incorrect. Marking a cell in the grid as non-walkable implicitly deletes whatever links that needs to delete, it cannot go wrong since there are no actual links to actually delete. This phenomenon still surprises me. Building a grid out of linked nodes is something I had never even considered doing before I saw someone else making that mistake. Surely a 2D array is the first thing you think of when your data has the shape of a 2D array, that just makes sen
For pathfinding, I just draw a bitmap in memory, at the lowest acceptable resolution, put all the obstacles on it, and draw a modified version of A* on top of that. Then I collect the result, and *bam* a path in < 100ms. The result usually gets stored on disk (as bitmap), which makes debugging waaaaay easier. This probably sounds god-awful for most people, but it's a really simple solution to a complex problem. Easy to maintain, easy to debug, easy to tweak on the fly (by modifying brush thickness of various obstacles!). Also, it avoids the common rookie mistake: trying to build an efficient solution to a problem you haven't solved yet. First you build the easiest possible solution, with considerations made towards debugging / testing / maintenance. Then you profile your resource consumption, so you know for a fact what's slow and what's not. After that, you refactor until you run out of time or budget to do so. EZPZ