I don't get the logic of nested list comprehensions in Python

nested_list_comprehension.py

""" The way nested list comprehensions work looks backwards to me.
    I use this pattern a lot to create a collection of DataFrames from stub data
    in writing tests for AWS Glue/PySpark ETL jobs.
"""
table_group = [
  {
    'database': 'ancient_legends_tier_1',
    'table_name': 'wizards',
    'columns': {
      'id': 'int',
      'name': 'string',
      'birth_epoch': 'int'
    },
    'rdd': [
      [ 1, 'Gandalf', 0 ],
      [ 2, 'Morgana', 3 ],
      [ 3, 'Galadriel', -10 ]
    ]
  },
  {
    'database': 'ancient_legends_tier_2',
    'table_name': 'epic_battles',
    'columns': {
      'id': 'int',
      'name': 'string',
      'wizard_id': 'int',
      'duration': 'float',
      'epicness': 'float'
    },
    'rdd': [
      [ 1, 'That Time Gandalf Almost Died In The Mines of Moria', 1, 3.2, 10 ],
      [ 2, 'Battle of Ealdor', 2, 0.125, 3 ]
    ]
  }
]

table_name = 'epic_battles'

# Intuitively, this seems to me like it should be
# col for col in t['columns'] if t['table_name'] == table_name for t in table_group
headings = [ col for t in table_group if t['table_name'] == table_name for col in t['columns'] ]