Make flare calculation go through all scores as PB may not have highest flare

[HutchyBen]

No description provided.

[beer-psi]

Fetching all scores and processing them in-memory is obscenely expensive. Other live-service games on Tachi with partitioned rating systems (e.g. old/new on GekiChuMai) have dropped in-game rating and only use NaiveRating. I think DDR should also drop in-game flare skill for a simple sum of best 90 play flare skills.

You can probably run the entire profile FlareSkill calculation as a MongoDB aggregation, though I'm not sure if that's better or worse

db.scores.aggregate([
    {
        $match: {
            game,
            playtype,
            userID,
            isPrimary: true,
            "calculatedData.flareSkill": {
                $type: "number"
            }
        }
    },
    {
        $group: {
            _id: "$chartID",
            songID: { $first: "$songID" }, // shouldn't be a problem since each chartID has only one songID
            flareSkill: { $max: "$calculatedData.flareSkill" }
        }
    },
    {
        $lookup: {
            from: "songs-ddr",
            localField: "songID",
            foreignField: "id",
            as: "song",
        }
    },
    {
        $set: {
            song: {
                $first: "$song"
            }
        }
    },
    {
        $sort: {
            "flareSkill": -1
        }
    },
    {
        $group: {
            _id: "$song.data.flareCategory",
            flareSkills: {
                $push: "$flareSkill"
            }
        }
    },
    {
        $set: {
            flareSkills: {
                $slice: ["$flareSkills", 30]
            }
        }
    },
    {
        $match: { _id: { $ne: "NONE" } }
    },
    {
        $group: {
            _id: null,
            flareSkill: {
                $sum: {
                    $sum: "$flareSkills"
                }
            }
        }
    }
])

This aggregation returns { flareSkill?: number; }. If we were on MongoDB 5.2+ we can replace the $sort, $group, $slice dance with a $topN accumulator, but I think we're on 5.0, since that's what the devcontainer is on.

{
  $group: {
    _id: "$song.data.flareCategory",
    flareSkills: {
      $topN: {
        n: 30,
        sortBy: {
          flareSkill: -1
        },
        output: "$flareSkill"
      }
    }
  }
}

[HutchyBen]

I initially thought there could be some sort of PB merge for best flare but I wasn't too sure how to go about that, I'll have another look now and if I cant figure it out i'll just close this PR and leave it as an issue for another time.

[beer-psi]

SDVX has a merge for best exScore, so you can probably just do that?

https://github.com/zkrising/Tachi/blob/4f3e154e7b58b8c1e861cb52dd33e76953e0fe36/server/src/game-implementations/games/sdvx.ts#L43-L45

[HutchyBen]

Yeah, I think when I was initailly looking at doing a PB merge I was trying to merge flareSkill and not the actual flare which is a bit silly thinking back now. I'll test it out merging the flare in a little bit