quicksight_bulk_update_datasets.py

#
# Update any Quicksight datasets with relational tables in the source schema to use the target schema.
#
# Run in dry run mode (Do this to test it)
# quicksight-bulk-update-datasets --account-id=<aws account id> --aws-profile=<aws profile name> --source-schema=<schema name> --target-schema=<schema name> --dry-run
#
# To run for real remove the "--dry-run" argument from the above command
#
# You will be prompted before each dataset is actually updated on Quicksight. To run it for all
# datasets without the prompt add the "--no-prompt" arg to the command
#
import csv
import datetime
from collections import deque
from contextlib import contextmanager
from typing_extensions import Annotated

import boto3
import pglast
import pglast.stream
import typer
from rich.highlighter import ReprHighlighter
from rich.progress import MofNCompleteColumn, BarColumn, SpinnerColumn, TextColumn, TimeRemainingColumn, Progress
from rich.table import Table
from botocore.exceptions import ClientError

app = typer.Typer()


@app.command()
def rename_schema(
        account_id: Annotated[str, typer.Option("--account-id", "-a", help="The AWS account ID", show_default=False)],
        source_schema: Annotated[str, typer.Option("--source-schema", "-s", help="The schema that will be renamed", show_default=False)],
        target_schema: Annotated[str, typer.Option("--target-schema", "-t", help="The new name of the source schema", show_default=False)],
        aws_profile: Annotated[str, typer.Option("--aws-profile", "-p", help="The profile to connect to AWS", show_default=False)] = None,
        dataset_id: Annotated[str, typer.Option("--dataset-id", "-i", help="Run for the dataset with this ID only")] = None,
        no_prompt: Annotated[bool, typer.Option("--no-prompt", "-n", help="Update all affected dataset without prompting the user")] = False,
        dry_run: Annotated[bool, typer.Option("--dry-run", "-d", help="Do not apply changes to Quicksight")] = False,
):
    """
    Update datasets that refer to the source schema to use the target schema
    """

    def all_dataset_ids(client):
        with Progress(SpinnerColumn(finished_text='[green]✔'), TextColumn("{task.description}")) as progress:
            task = progress.add_task("Finding all datasets...", total=1)

            dataset_ids = tuple(
                dataset_summary['DataSetId']
                for page in client.get_paginator("list_data_sets").paginate(AwsAccountId=account_id)
                for dataset_summary in page.get("DataSetSummaries", [])
            )
            progress.update(task, advance=1, description=f'Finding all datasets... done: [bold]{len(dataset_ids)}')
            return dataset_ids

    def datasets(client, console, dataset_ids):
        for _dataset_id in dataset_ids:
            try:
                response = client.describe_data_set(
                    AwsAccountId=account_id, DataSetId=_dataset_id
                )
            except ClientError as ex:
                # Some datasets cannot be described in the API, e.g. manual uploads
                console.print(f"⚠️  Error fetching detail for dataset {_dataset_id}: {ex.response['Error']['Message']}")
                continue

            yield response["DataSet"]

    @contextmanager
    def csv_output_report():
        timestamp = datetime.datetime.now().strftime('%Y%m%d-%H%M%S')
        filename = f"{timestamp}--{account_id}{'--dry-run' if dry_run else ''}.csv"

        with open(filename, 'w', newline='') as f:
            fieldnames = ['dataset_id', 'dataset_link', 'physical_table_id', 'type', 'source', 'target', 'error']
            writer = csv.DictWriter(f, fieldnames=fieldnames)
            writer.writeheader()

            yield writer

    def tables_from_query(query):
        statements = pglast.parse_sql(query)

        if len(statements) == 0:
            return []

        tables = set()

        node_ctenames = deque()
        node_ctenames.append((statements[0](), ()))

        while node_ctenames:
            node, ctenames = node_ctenames.popleft()

            if node.get("withClause", None) is not None:
                if node["withClause"]["recursive"]:
                    ctenames += tuple((cte["ctename"] for cte in node["withClause"]["ctes"]))
                    for cte in node["withClause"]["ctes"]:
                        node_ctenames.append((cte, ctenames))
                else:
                    for cte in node["withClause"]["ctes"]:
                        node_ctenames.append((cte, ctenames))
                        ctenames += (cte["ctename"],)

            if node.get("@", None) == "RangeVar" and (
                    node["schemaname"] is not None or node["relname"] not in ctenames
            ):
                tables.add((node["schemaname"] or "public", node["relname"]))

            for node_type, node_value in node.items():
                if node_type == "withClause":
                    continue
                for nested_node in node_value if isinstance(node_value, tuple) else (node_value,):
                    if isinstance(nested_node, dict):
                        node_ctenames.append((nested_node, ctenames))

        return sorted(list(tables))

    def rename_schema(query, source, target):
        # Strongly based on tables_from_query, but that is based on iterating over the "dict" form
        # of the parsed query. But to re-serialize, we need the object form. Could probably be
        # simplified to only use the object form
        statements = pglast.parse_sql(query)

        if len(statements) > 1:
            raise Exception("Multiple statements are not supported")

        if len(statements) == 0:
            return []

        node_ctenames = deque()
        node_ctenames.append((statements[0], ()))

        while node_ctenames:
            node, ctenames = node_ctenames.popleft()
            node_dict = node()

            if node_dict.get("withClause", None) is not None:
                if node_dict["withClause"]["recursive"]:
                    ctenames += tuple((cte.ctename for cte in node.withClause.ctes))
                    for cte in node.withClause.ctes:
                        node_ctenames.append((cte, ctenames))
                else:
                    for cte in node.withClause.ctes:
                        node_ctenames.append((cte, ctenames))
                        ctenames += (cte.ctename,)

            if node_dict.get("@", None) == "RangeVar" and (
                node_dict["schemaname"] is not None or node_dict["relname"] not in ctenames
            ) and (node.schemaname or "public") == source:
                node.schemaname = target
            if (node_dict.get("@", None) == "ColumnRef" and
                len(node.fields) == 3 and
                node.fields[0].sval == source
            ):
                node.fields[0].sval = target

            for node_type in node:
                node_value = getattr(node, node_type)
                if node_type == "withClause":
                    continue
                for nested_node in node_value if isinstance(node_value, tuple) else (node_value,):
                    if isinstance(nested_node, pglast.ast.Node):
                        node_ctenames.append((nested_node, ctenames))

        return pglast.stream.IndentedStream(
            comments=pglast._extract_comments(query),
            comma_at_eoln=True,
        )(statements[0])

    def modify_dataset_dict_if_needed(console, dataset):
        for physical_table_id, physical_table in dataset.get("PhysicalTableMap", {}).items():
            dataset_link = f"https://eu-west-2.quicksight.aws.amazon.com/sn/data-sets/{dataset['DataSetId']}"
            if "RelationalTable" in physical_table:
                table = physical_table["RelationalTable"]
                if table["Schema"] == source_schema:
                    grid = Table.grid(padding=(0, 1))
                    grid.add_column(justify="left")
                    grid.add_column(justify="left")
                    grid.add_row(f"[green]✔[/green] RelationalTable", ReprHighlighter()(f"{(table['Schema'], table['Name'])} ➜ {(target_schema, table['Name'])}"))
                    console.print(grid, highlight=True)
                    table["Schema"] = target_schema
                    yield {
                        'dataset_id': dataset['DataSetId'],
                        'dataset_link': dataset_link,
                        'physical_table_id': physical_table_id,
                        'type': 'table',
                        'source': source_schema,
                        'target': target_schema,
                        'error': '',
                    }
            if "CustomSql" in physical_table:
                original_sql = physical_table["CustomSql"]["SqlQuery"]
                try:
                    tables = tables_from_query(physical_table["CustomSql"]["SqlQuery"])
                except Exception as e:
                    console.print(f"⚠️  Unable to parse query in {dataset['DataSetId']}")
                    yield {
                        'dataset_id': dataset['DataSetId'],
                        'dataset_link': dataset_link,
                        'physical_table_id': physical_table_id,
                        'type': 'sql',
                        'source': original_sql,
                        'target': '',
                        'error': f'Unable to parse query: {e}',
                    }
                    continue
                if any(schema == source_schema for schema, table in tables):
                    renamed_sql = rename_schema(original_sql, source_schema, target_schema)
                    new_tables = tables_from_query(renamed_sql)
                    grid = Table.grid(padding=(0, 1))
                    grid.add_column(justify="left")
                    grid.add_column(justify="left")
                    innerGrid = Table.grid(padding=(0, 1))
                    innerGrid.add_column(justify="left")
                    innerGrid.add_column(justify="left")
                    for old, new in zip(tables, new_tables):
                        innerGrid.add_row(ReprHighlighter()(f'{old}'), '➜', ReprHighlighter()(f'{new}'))
                    grid.add_row(f"[green]✔[/green] CustomSql      ", innerGrid)
                    console.print(grid, highlight=True)
                    no_original = any(schema == source_schema for schema, table in new_tables) == False
                    any_target = any(schema == target_schema for schema, table in new_tables) == True
                    original_tables_to_be_unchanged = set((schema, table) for schema, table in tables if schema not in (source_schema, target_schema))
                    new_tables_to_be_unchanged = set((schema, table) for schema, table in new_tables if schema not in (source_schema, target_schema))
                    assert no_original
                    assert any_target
                    assert original_tables_to_be_unchanged == new_tables_to_be_unchanged
                    physical_table["CustomSql"]["SqlQuery"] = renamed_sql
                    yield {
                        'dataset_id': dataset['DataSetId'],
                        'dataset_link': dataset_link,
                        'physical_table_id': physical_table_id,
                        'type': 'sql',
                        'source': original_sql,
                        'target': rename_schema(original_sql, source_schema, target_schema),
                        'error': '',
                    }

    session_opts = {'profile_name': aws_profile} if aws_profile is not None else {}
    session = boto3.Session(**session_opts)
    client = session.client("quicksight")

    dataset_ids = \
        (dataset_id,) if dataset_id is not None else \
        all_dataset_ids(client)

    updated = 0

    with \
            csv_output_report() as writer, \
            Progress(
                SpinnerColumn(finished_text='[green]✔'),
                TextColumn("[progress.description]{task.description}"),
                BarColumn(),
                MofNCompleteColumn(),
                '[bright_yellow]Updated: {task.fields[updated]}',
                TimeRemainingColumn(),
            ) as progress:

        task = progress.add_task(("[green]\\[DRY RUN][/green] " if dry_run else "") + 'Updating datasets...', total=len(dataset_ids), updated=0)
        
        for dataset in datasets(client, progress.console, dataset_ids):

            dataset_changes = tuple(modify_dataset_dict_if_needed(progress.console, dataset))

            non_errored_changes = [not dataset_change['error'] for dataset_change in dataset_changes]

            if non_errored_changes and not dry_run:
                if not no_prompt:
                    input("Press enter to update the dataset on Quicksight")

                try:
                    client.update_data_set(
                        AwsAccountId=account_id,
                        **{
                            x: v
                            for x, v in dataset.items()
                            if x
                            not in [
                                "Arn",
                                "CreatedTime",
                                "LastUpdatedTime",
                                "OutputColumns",
                                "ConsumedSpiceCapacityInBytes",
                            ]
                        },
                    )
                except Exception as e:
                    progress.console.print(f"⚠️  Failed to update dataset: {e}")
                    for dataset_change in dataset_changes:
                        dataset_change['error'] = f'Failed to update dataset: {e}'

            for dataset_change in dataset_changes:
                writer.writerow(dataset_change)

            non_errored_changes_post_update = [not dataset_change['error'] for dataset_change in dataset_changes]

            updated += 1 if non_errored_changes_post_update else 0
            progress.advance(task)
            progress.update(task, updated=updated)