[YouTube] Improve lockup view model metadata extraction

Nobody has claimed this yet.

The metadata in the lockup view model is distributed across 2/3 rows and possibly there are multiple items per row. As of #1320 the YouTubeStreamInfoItemLockupExtractor accesses specific rows and columns by index, expecting to always find the same piece of data (uploader name, view count, upload date) in the same place.

The following was originally posted by @Stypox in https://github.com/TeamNewPipe/NewPipeExtractor/pull/1320#discussion_r2205778179

Can we trust the indices to be stable? I guess we can't do much better anyway, so let's try like this for now. If it creates problems, we can always use other ways to detect the right metadata parts in the future, e.g.:

the uploader name object has "attachmentRuns" and "styleRuns" attached to it, which view count and upload date don't have
the view count contains the "views" string or the "watching" string (for lives)
the remaining item is the upload date

See here

[
  {
    "metadataParts": [
      {
        "text": {
          "content": "Justin Bieber",
          "styleRuns": [
            {
              "startIndex": 13,
              "styleRunExtensions": {
                "styleRunColorMapExtension": {
                  "colorMap": [
                    {
                      "key": "USER_INTERFACE_THEME_LIGHT",
                      "value": 4284506208
                    },
                    {
                      "key": "USER_INTERFACE_THEME_DARK",
                      "value": 4289374890
                    }
                  ]
                }
              }
            }
          ],
          "attachmentRuns": [
            {
              "startIndex": 13,
              "length": 0,
              "element": {
                "type": {
                  "imageType": {
                    "image": {
                      "sources": [
                        {
                          "clientResource": {
                            "imageName": "AUDIO_BADGE"
                          },
                          "width": 14,
                          "height": 14
                        }
                      ]
                    }
                  }
                },
                "properties": {
                  "layoutProperties": {
                    "height": {
                      "value": 14,
                      "unit": "DIMENSION_UNIT_POINT"
                    },
                    "width": {
                      "value": 14,
                      "unit": "DIMENSION_UNIT_POINT"
                    },
                    "margin": {
                      "left": {
                        "value": 4,
                        "unit": "DIMENSION_UNIT_POINT"
                      }
                    }
                  }
                }
              },
              "alignment": "ALIGNMENT_VERTICAL_CENTER"
            }
          ]
        }
      }
    ]
  },
  {
    "metadataParts": [
      {
        "text": {
          "content": "157M views"
        }
      },
      {
        "text": {
          "content": "5 years ago"
        }
      }
    ]
  },
  {
    "badges": [
      {
        "badgeViewModel": {
          "badgeText": "Free with ads",
          "badgeStyle": "BADGE_COMMERCE",
          "trackingParams": "COcBENTEDBgBIhMIs6r8_ZS9jgMVOxYGAB1LCied"
        }
      },
      {
        "badgeViewModel": {
          "badgeText": "T",
          "badgeStyle": "BADGE_DEFAULT",
          "trackingParams": "COcBENTEDBgBIhMIs6r8_ZS9jgMVOxYGAB1LCied"
        }
      }
    ]
  }
]

Start with YouTubeStreamInfoItemLockupExtractor and inspect how it reads metadata rows and columns from the lockup view-model response. Use the JSON example in this issue to compare the uploader, view-count, and upload-date entries, then make the extraction identify those parts reliably rather than relying only on fixed indices. Done means the three metadata values remain correctly identified when rows contain multiple items or change order.

Written by the indexing model from the issue text.

Description

Contributor guide

First steps

Research direction

Assessment