Show parsing state in the playground. When dropping fields for example

[hellovai]

The rows field is dropped: TODO figure out why.

https://www.promptfiddle.com/Parse-Image-UTWNb

Here is the extracted data from the table in JSON format:

```json
{
  "headers": [
    "No. of establishments (1963.12.31)",
    "No. of workers (1963.12.31)",
    "Output (1963.1.1~1963.12.31)",
    "Value of shipments (1963.1.1~1963.12.31)",
    "Production costs (1963.1.1~1963.12.31)",
    "Value added (1963.1.1~1963.12.31)"
  ],
  "columns": [
    "Total",
    "5-9",
    "10-19",
    "20-29",
    "30-49",
    "50-99",
    "100-199",
    "200-499",
    "500 and above"
  ],
  "rows": [
    [18310, 401981, 166857471, 163199452, 105323904, 61533567],
    [10132, 60564, 16925825, 16759437, 10939674, 5986151],
    [4581, 36038, 18871885, 18920639, 11906389, 6832271],
    [1387, 30500, 10187044, 9930658, 6836101, 3319350],
    [965, 34258, 12345164, 12210238, 8168539, 4216625],
    [713, 46811, 20317965, 21317448, 13862461, 6019839],
    [295, 38651, 19058913, 19597913, 12877183, 6719730],
    [165, 45767, 22734463, 22667817, 15796718, 6861489],
    [72, 89932, 46463544, 45162100, 26072300, 80258480]
  ]
}
\(remove <-this last \)```

Parsed:

{
headers: [
0: "No. of establishments (1963.12.31)"
1: "No. of workers (1963.12.31)"
2: "Output (1963.1.1~1963.12.31)"
3: "Value of shipments (1963.1.1~1963.12.31)"
4: "Production costs (1963.1.1~1963.12.31)"
5: "Value added (1963.1.1~1963.12.31)"
]
columns: [
0: "Total"
1: "5-9"
2: "10-19"
3: "20-29"
4: "30-49"
5: "50-99"
6: "100-199"
7: "200-499"
8: "500 and above"
]
row: [
]
}

class Table {
  headers string[]
  columns string[]
  row int[][] @description(#"
    The rows of the table, each containing the data for each column.
  "#)
}

[hellovai]

This issue is not a parsing failure, but rather the fact that the field had the word row but the llm output rows. We should consider exposing better errors up.

In this case we successfully parsed, but only because lists have a default value of an empty list when not found.

[aaronvg]

omg this couldve been caught by an AI linter (or even a programmatic one)