Best way to load questslog.txt

[Veracity]

Due to difficulties in parsing a many-dimensional map into 2D. Wrote a nifty recursive function to do it, but ASH got flummoxed by my attempt to foreach an "aggregate".

By "got flummoxed" I assume you mean "did not have syntax to represent what I wanted to do."

Show me your function and lets talk about it. We had a discussion years about about letting functions deal with generic aggregates, much as map_to_file and file_to_map do. Perhaps this is more of the same. Perhaps it is time to revisit it and decide if there is anything we can do to the syntax that makes sense that would make your life easier.

Perhaps that should go into its own "scripting discussion" thread.

[heeheehee]

I'm wondering if this problem could be solved by adding a pair of functions that operate similarly to map_to_file/file_to_map.

Namely, two possibilities readily come to mind:

1. Treat the entire file as tab-separated values with no additional structure -- works specifically with type[int, int].

2. Treat the file as a map, but in each line, interpret fields from 2 onward as an array -- so, type_v[type_k, int].

[zarqon]

@Darzil: Thanks!

By "got flummoxed" I assume you mean "did not have syntax to represent what I wanted to do."

Quite so. I hadn't expected it to work, honestly, with ASH being typed and all, but since I also recalled that discussion long ago which resulted in allowing passing a generic "aggregate" to a function, I figured it was worth a shot. I've since deleted it, but I can try to reconstruct it here. It was something like

void onerow(aggregate a, int depth) {
   if (depth > 100) return;
   foreach key, value in a {       // this is where the error was thrown
      add_table_row(depth, key);   // adds another <tr> to an HTML table being built in a buffer
      if (some test of value to determine it is an aggregate rather than a string) 
         onerow(value, depth + 1);
   }
}

[Veracity]

2. Treat the file as a map, but in each line, interpret fields from 2 onward as an array -- so, type_v[type_k, int].

I'm thinking we almost have something like this already, given that I added arrays of unspecified length.

typedef string[] type_v;
typedef string type_k;
type_v [type_k] quests;
file_to_map( "questlogs.txt", quests );

That almost certainly doesn't work, currently.

It's also probably not quite right, because file_to_map insists on using an existing map, rather than crafting and returning a new one. It does that because the passed in map defines the type of map.

Huh. Actually, it might be right; the "value" type is a length-0 array, which is defined to be "an array which will be initialized to the actual number of elements given to it, but which is immutable afterwords."

To make that work CompositeValue.read() would have to recognize that if the datatype is a zero-length array, it should simply consume all the remaining values on the line and initialize an array of the correct length as the value. Perhaps by defining ArrayValue.read() to understand this case.

[heeheehee]

To make that work CompositeValue.read() would have to recognize that if the datatype is a zero-length array, it should simply consume all the remaining values on the line and initialize an array of the correct length as the value. Perhaps by defining ArrayValue.read() to understand this case.

This would be interesting, but I'm somewhat concerned that this might be backwards-incompatible. That said, I feel like anyone who's using new features like 0-length arrays should expect some degree of churn (and keep an eye out for this sort of thing).

[Veracity]

With a bunch of changes to KoLmafia, this script:

typedef string[] type_v;
typedef string type_k;
type_v [type_k] quests;
file_to_map( "questslog.txt", quests );

print( "There are " + count( quests ) + " quests." );

int longest_quest = 0;
string key = "";
foreach quest, steps in quests {
    if ( count( steps ) > longest_quest ) {
	longest_quest = count( steps );
	key = quest;
    }
}

print( "Quest '" + key + "' has " + longest_quest + " steps." );

type_v steps = quests[ key ];
print( steps );

foreach i, step in steps {
    print( i + ": " + step );
}

yields this:

> Test/questlog.ash

There are 96 quests.
Quest 'questG04Nemesis' has 32 steps.
aggregate string [32]
0: Me and My Nemesis
1: Search the Misspelled Cemetary on the Nearby Plains for the Tomb of the Unknown
2: in the Misspelled Cemetary wants to fight you to test your worthiness.
3: Talk to the Ghost of the Unknown
4: Retrieve the Epic Weapon!
5: You've reclaimed the Epic Weapon! Time to go show it off at your guild.
6: recover the missing piece of the Legendary Epic Weapon.
7: Inform your guild that Beelzebozo has been dispatched.
8: Meatsmith the two parts of the Legendary Epic Weapon back together -- you'll need a tenderizing hammer from the Meatsmith.
9: Take the Legendary Epic Weapon back to your guild.
10: (no quest entry at all!)
11: The hunt for your Nemesis is on! Better check out that cave they sent you to.
12: Figure out how to get into your Nemesis' cave. If you're stumped, maybe your guild can help?
13: The cavern is full of weird mushrooms, but where's your Nemesis?
14: more fizzing spore pods to blow up the blockade in your Nemesis' cave.
15: Take those fizzing spore pods to the rubble!
16: Boom! Time to bring the fight to your stinking Nemesis in that stinking cave!
17: Heck yeah, you beat your Nemesis and got a sweet hat. Better take it back to your guild.
18: You're waiting for your guild's scouts to find out where your Nemesis went.
19: It appears as though some nefarious ne'er-do-well has put a contract on your head!<p>Gee, I wonder who it could be...
20: You handily dispatched some thugs who were trying to collect on your bounty, but something tells you they won't be the last ones to try.
21: Whoever put this hit out on you (like you haven't guessed already) has sent Mob Penguins to do their dirty work. Do you know any polar bears you could hire as bodyguards? No? Looks like you're on your own, then.
22: So much for those mob penguins that were after your head! If whoever put this hit out on you wants you killed (which, presumably, they do) they'll have to find some much more competent thugs.
23: Your suspicious[sic] have been confirmed: your Nemesis has put the order out for you to be hunted down and killed, and now they're sending their own guys instead of contracting out. Good luck!
24: Bam! So much for your Nemesis' assassins! If that's the best they've got, you have nothing at all to worry about.<p>You sure hope that's the best they've got.
25: You had a run-in with some crazy mercenary or assassin or... thing that your Nemesis sent to do you in once and for all. A run-in followed by a run-out, evidently, but you're going to have to deal with this sooner or later.
26: Now that you've dealt with your Nemesis' assassins and found a map to the secret tropical island volcano lair, it's time to take the fight to your foe. Booyah.
27: You've arrived at the secret tropical island volcano lair, and it's time to finally put a stop to this Nemesis nonsense once and for all. As soon as you can find where they're hiding. Maybe you can find someone to ask.
28: got away. Again.
29: Congratulations on solving the lava maze, which is probably the biggest pain-in-the-ass puzzle in the entire game! Hooray! (Unless you cheated, in which case: Boo!)
30: have some sort of crazy powerful and hideous final form? I was, but then I wrote all of this, so, y'know.
31: Okay, now you've defeated your Nemesis once and for all! (Well, probably. Almost definitely.) Nevermore will the members of whichever class you're playing at the moment be hassled by that terrible entity! Way to go!

Issues I found with aggregate literals:

1) They did not work in foreach since they didn't have iterators
2) Given iterators, you couldn't fetch values, since they didn't have aref
[3) They didn't have aset either. Since they are immutable, they should, but should generate a runtime error]
4) count() always returned 0

I fixed all of the above and also modified file_to_map as proposed: if you read into a 0-length array, you get all the remaining fields on the line read as the appropriate data type.

This is backwards-incompatible since if you tried it before, we threw an exception, caught it, and reported a "bad line" in your input file. Which is to say, it never did anything reasonable before and now it does.

As coded, it only works if your type_v is a simple type - one field per value. If you want the remaining fields to be <n> records, each with three fields, thus consuming fields three at a time, no go. I'd have to see an actual use case for this before I'd be inclined to make that work.

[Veracity]

(Revision 18040)

[Veracity]

Hm. I did it via an ArrayLiteral since the code already existed for reading a variable number of elements into an array[]. I wonder if it might be better to duplicate that code and get an ArrayValue, not an ArrayLiteral?

I will experiment.

[Veracity]

Yes, that works nicely.

Further experimentation showed that we couldn't use map_file to write out the map in the same format, however. That is because arrays do not have a "compact" representation in data files, unlike records.

I have an implementation of "compact" file format for arrays which works nicely, but which is non-backwards compatible without a small code change in scripts: if you have a map whose values are arrays - not maps! - and did map_to_file, it wrote it out in non-compact form - even though the default for map_to_file is compact. The solution is to change your map_to_file and file_to_map calls to specify "false" for the "is this compact" format.

I'll create a new thread to discuss this.

[Bale]

Just thought I'd let you know that I'm now making use of this feature in OCD. It helps when reading concoctions.txt since there are a variable number of ingredients for a given concoction. The old implementation worked, just as long as no new super long concoctions exceeded my dataspace. I worried that one day it would fail when KOL implemented something new.

zarqon isn't the only one making use of your work.

Thanks!