Flowfile 0.3.3.2__py3-none-any.whl → 0.3.4.1__py3-none-any.whl

flowfile_core/flowfile/code_generator/code_generator.py

@@ -0,0 +1,723 @@
+from typing import List, Dict, Optional, Set, Tuple, Any
+from collections import defaultdict
+import polars as pl
+
+from flowfile_core.flowfile.flow_graph import FlowGraph
+from flowfile_core.flowfile.flow_data_engine.flow_file_column.main import FlowfileColumn, convert_pl_type_to_string
+from flowfile_core.flowfile.flow_data_engine.flow_file_column.utils import cast_str_to_polars_type
+from flowfile_core.flowfile.flow_node.flow_node import FlowNode
+from flowfile_core.flowfile.util.execution_orderer import determine_execution_order
+from flowfile_core.schemas import input_schema, transform_schema
+from flowfile_core.configs import logger
+
+
+class FlowGraphToPolarsConverter:
+    """
+    Converts a FlowGraph into executable Polars code.
+
+    This class takes a FlowGraph instance and generates standalone Python code
+    that uses only Polars, without any Flowfile dependencies.
+    """
+    flow_graph: FlowGraph
+    node_var_mapping: Dict[int, str]
+    imports: Set[str]
+    code_lines: List[str]
+    output_nodes: List[Tuple[int, str]] = []
+    last_node_var: Optional[str] = None
+
+    def __init__(self, flow_graph: FlowGraph):
+        self.flow_graph = flow_graph
+        self.node_var_mapping: Dict[int, str] = {}  # Maps node_id to variable name
+        self.imports: Set[str] = {"import polars as pl"}
+        self.code_lines: List[str] = []
+        self.output_nodes = []
+        self.last_node_var = None
+
+    def convert(self) -> str:
+        """
+        Main method to convert the FlowGraph to Polars code.
+
+        Returns:
+            str: Complete Python code that can be executed standalone
+        """
+        # Get execution order
+        execution_order = determine_execution_order(
+            all_nodes=[node for node in self.flow_graph.nodes if node.is_correct],
+            flow_starts=self.flow_graph._flow_starts + self.flow_graph.get_implicit_starter_nodes()
+        )
+
+        # Generate code for each node in order
+        for node in execution_order:
+            self._generate_node_code(node)
+
+        # Combine everything
+        return self._build_final_code()
+
+    def handle_output_node(self, node: FlowNode, var_name: str) -> None:
+        settings = node.setting_input
+        if hasattr(settings, 'is_flow_output') and settings.is_flow_output:
+            self.output_nodes.append((node.node_id, var_name))
+
+    def _generate_node_code(self, node: FlowNode) -> None:
+        """Generate Polars code for a specific node."""
+        node_type = node.node_type
+        settings = node.setting_input
+        # Skip placeholder nodes
+        if isinstance(settings, input_schema.NodePromise):
+            self._add_comment(f"# Skipping uninitialized node: {node.node_id}")
+            return
+        # Create variable name for this node's output
+        var_name = f"df_{node.node_id}"
+        self.node_var_mapping[node.node_id] = var_name
+        self.handle_output_node(node, var_name)
+        if node.node_template.output>0:
+            self.last_node_var = var_name
+        # Get input variable names
+        input_vars = self._get_input_vars(node)
+        # Route to appropriate handler based on node type
+        handler = getattr(self, f"_handle_{node_type}", None)
+        if handler:
+            handler(settings, var_name, input_vars)
+        else:
+            self._add_comment(f"# TODO: Implement handler for node type: {node_type}")
+            raise Exception(f"No handler implemented for node type: {node_type}")
+
+    def _get_input_vars(self, node: FlowNode) -> Dict[str, str]:
+        """Get input variable names for a node."""
+        input_vars = {}
+
+        if node.node_inputs.main_inputs:
+            if len(node.node_inputs.main_inputs) == 1:
+                input_vars['main'] = self.node_var_mapping.get(
+                    node.node_inputs.main_inputs[0].node_id, 'df'
+                )
+            else:
+                for i, input_node in enumerate(node.node_inputs.main_inputs):
+                    input_vars[f'main_{i}'] = self.node_var_mapping.get(
+                        input_node.node_id, f'df_{i}'
+                    )
+
+        if node.node_inputs.left_input:
+            input_vars['left'] = self.node_var_mapping.get(
+                node.node_inputs.left_input.node_id, 'df_left'
+            )
+
+        if node.node_inputs.right_input:
+            input_vars['right'] = self.node_var_mapping.get(
+                node.node_inputs.right_input.node_id, 'df_right'
+            )
+
+        return input_vars
+
+    def _handle_csv_read(self, file_settings: input_schema.ReceivedTable, var_name: str):
+        if file_settings.encoding.lower() in ('utf-8', 'utf8'):
+            encoding = "utf8-lossy"
+            self._add_code(f"{var_name} = pl.scan_csv(")
+            self._add_code(f'    "{file_settings.abs_file_path}",')
+            self._add_code(f'    separator="{file_settings.delimiter}",')
+            self._add_code(f'    has_header={file_settings.has_headers},')
+            self._add_code(f'    ignore_errors={file_settings.ignore_errors},')
+            self._add_code(f'    encoding="{encoding}",')
+            self._add_code(f'    skip_rows={file_settings.starting_from_line},')
+            self._add_code(")")
+        else:
+            self._add_code(f"{var_name} = pl.read_csv(")
+            self._add_code(f'    "{file_settings.abs_file_path}",')
+            self._add_code(f'    separator="{file_settings.delimiter}",')
+            self._add_code(f'    has_header={file_settings.has_headers},')
+            self._add_code(f'    ignore_errors={file_settings.ignore_errors},')
+            if file_settings.encoding:
+                self._add_code(f'    encoding="{file_settings.encoding}",')
+            self._add_code(f'    skip_rows={file_settings.starting_from_line},')
+            self._add_code(").lazy()")
+
+    def _handle_read(self, settings: input_schema.NodeRead, var_name: str, input_vars: Dict[str, str]) -> None:
+        """Handle file reading nodes."""
+        file_settings = settings.received_file
+
+        if file_settings.file_type == 'csv':
+            self._handle_csv_read(file_settings, var_name)
+
+        elif file_settings.file_type == 'parquet':
+            self._add_code(f'{var_name} = pl.scan_parquet("{file_settings.abs_file_path}")')
+
+        elif file_settings.file_type in ('xlsx', 'excel'):
+            self._add_code(f"{var_name} = pl.read_excel(")
+            self._add_code(f'    "{file_settings.abs_file_path}",')
+            if file_settings.sheet_name:
+                self._add_code(f'    sheet_name="{file_settings.sheet_name}",')
+            self._add_code(").lazy()")
+
+        self._add_code("")
+
+    @staticmethod
+    def _generate_pl_schema_with_typing(flowfile_schema: List[FlowfileColumn]) -> str:
+        polars_schema_str = "pl.Schema([" + ", ".join(f'("{flowfile_column.column_name}", pl.{flowfile_column.data_type})'
+                                   for flowfile_column in flowfile_schema) + "])"
+        return polars_schema_str
+
+    def get_manual_schema_input(self, flowfile_schema: List[FlowfileColumn]) -> str:
+        polars_schema_str = self._generate_pl_schema_with_typing(flowfile_schema)
+        is_valid_pl_schema = self._validate_pl_schema(polars_schema_str)
+        if is_valid_pl_schema:
+            return polars_schema_str
+        else:
+            return "[" + ", ".join([f'"{c.name}"' for c in flowfile_schema]) + "]"
+
+    @staticmethod
+    def _validate_pl_schema(pl_schema_str: str) -> bool:
+        try:
+            _globals = {"pl": pl}
+            eval(pl_schema_str, _globals)
+            return True
+        except Exception as e:
+            logger.error(f"Invalid Polars schema: {e}")
+            return False
+
+    def _handle_manual_input(self, settings: input_schema.NodeManualInput, var_name: str, input_vars: Dict[str, str]) -> None:
+        """Handle manual data input nodes."""
+        if settings.raw_data_format:
+            data = settings.raw_data_format.data
+            flowfile_schema = list(FlowfileColumn.create_from_minimal_field_info(c) for c in settings.raw_data_format.columns)
+            schema = self.get_manual_schema_input(flowfile_schema)
+            self._add_code(f"{var_name} = pl.LazyFrame({data}, schema={schema}, strict=False)")
+        else:
+            self._add_code(f"{var_name} = pl.LazyFrame({settings.raw_data})")
+        self._add_code("")
+
+    def _handle_filter(self, settings: input_schema.NodeFilter, var_name: str, input_vars: Dict[str, str]) -> None:
+        """Handle filter nodes."""
+        input_df = input_vars.get('main', 'df')
+
+        if settings.filter_input.filter_type == 'advanced':
+            # Parse the advanced filter expression
+            self.imports.add(
+                "from polars_expr_transformer.process.polars_expr_transformer import simple_function_to_expr"
+            )
+            self._add_code(f"{var_name} = {input_df}.filter(")
+            self._add_code(f'simple_function_to_expr("{settings.filter_input.advanced_filter}")')
+            self._add_code(")")
+        else:
+            # Handle basic filter
+            basic = settings.filter_input.basic_filter
+            filter_expr = self._create_basic_filter_expr(basic)
+            self._add_code(f"{var_name} = {input_df}.filter({filter_expr})")
+        self._add_code("")
+
+    def _handle_record_count(self, settings: input_schema.NodeRecordCount, var_name: str, input_vars: Dict[str, str]):
+        input_df = input_vars.get('main', 'df')
+        self._add_code(f"{var_name} = {input_df}.select(pl.len().alias('number_of_records'))")
+
+    def _handle_graph_solver(self, settings: input_schema.NodeGraphSolver, var_name: str, input_vars: Dict[str, str]):
+        input_df = input_vars.get('main', 'df')
+        from_col_name = settings.graph_solver_input.col_from
+        to_col_name = settings.graph_solver_input.col_to
+        output_col_name = settings.graph_solver_input.output_column_name
+        self._add_code(f'{var_name} = {input_df}.with_columns(graph_solver(pl.col("{from_col_name}"), '
+                       f'pl.col("{to_col_name}"))'
+                       f'.alias("{output_col_name}"))')
+        self._add_code("")
+        self.imports.add("from polars_grouper import graph_solver")
+
+    def _handle_select(self, settings: input_schema.NodeSelect, var_name: str, input_vars: Dict[str, str]) -> None:
+        """Handle select/rename nodes."""
+        input_df = input_vars.get('main', 'df')
+        # Get columns to keep and renames
+        select_exprs = []
+        for select_input in settings.select_input:
+            if select_input.keep and select_input.is_available:
+                if select_input.old_name != select_input.new_name:
+                    expr = f'pl.col("{select_input.old_name}").alias("{select_input.new_name}")'
+                else:
+                    expr = f'pl.col("{select_input.old_name}")'
+
+                if (select_input.data_type_change or select_input.is_altered) and select_input.data_type:
+                    polars_dtype = self._get_polars_dtype(select_input.data_type)
+                    expr = f'{expr}.cast({polars_dtype})'
+
+                select_exprs.append(expr)
+
+        if select_exprs:
+            self._add_code(f"{var_name} = {input_df}.select([")
+            for expr in select_exprs:
+                self._add_code(f"    {expr},")
+            self._add_code("])")
+        else:
+            self._add_code(f"{var_name} = {input_df}")
+        self._add_code("")
+
+    def _handle_join(self, settings: input_schema.NodeJoin, var_name: str, input_vars: Dict[str, str]) -> None:
+        """Handle join nodes."""
+        left_df = input_vars.get('main', input_vars.get('main_0', 'df_left'))
+        right_df = input_vars.get('right', input_vars.get('main_1', 'df_right'))
+
+        # Extract join keys
+        left_on = [jm.left_col for jm in settings.join_input.join_mapping]
+        right_on = [jm.right_col for jm in settings.join_input.join_mapping]
+
+        self._add_code(f"{var_name} = {left_df}.join(")
+        self._add_code(f"    {right_df},")
+        self._add_code(f"    left_on={left_on},")
+        self._add_code(f"    right_on={right_on},")
+        self._add_code(f'    how="{settings.join_input.how}"')
+        self._add_code(")")
+        self._add_code("")
+
+    def _handle_group_by(self, settings: input_schema.NodeGroupBy, var_name: str, input_vars: Dict[str, str]) -> None:
+        """Handle group by nodes."""
+        input_df = input_vars.get('main', 'df')
+
+        # Separate groupby columns from aggregation columns
+        group_cols = []
+        agg_exprs = []
+
+        for agg_col in settings.groupby_input.agg_cols:
+            if agg_col.agg == 'groupby':
+                group_cols.append(agg_col.old_name)
+            else:
+                agg_func = self._get_agg_function(agg_col.agg)
+                expr = f'pl.col("{agg_col.old_name}").{agg_func}().alias("{agg_col.new_name}")'
+                agg_exprs.append(expr)
+
+        self._add_code(f"{var_name} = {input_df}.group_by({group_cols}).agg([")
+        for expr in agg_exprs:
+            self._add_code(f"    {expr},")
+        self._add_code("])")
+        self._add_code("")
+
+    def _handle_formula(self, settings: input_schema.NodeFormula, var_name: str, input_vars: Dict[str, str]) -> None:
+        """Handle formula/expression nodes."""
+        input_df = input_vars.get('main', 'df')
+        self.imports.add("from polars_expr_transformer.process.polars_expr_transformer import simple_function_to_expr")
+
+        # Convert SQL-like formula to Polars expression
+        formula = settings.function.function
+        col_name = settings.function.field.name
+        self._add_code(f"{var_name} = {input_df}.with_columns([")
+        self._add_code(f'simple_function_to_expr({repr(formula)}).alias("{col_name}")')
+        if settings.function.field.data_type not in (None, "Auto"):
+            output_type = convert_pl_type_to_string(cast_str_to_polars_type(settings.function.field.data_type))
+            if output_type[:3] != "pl.":
+                output_type = "pl." + output_type
+            self._add_code(f'    .cast({output_type})')
+
+        self._add_code("])")
+        self._add_code("")
+
+    def _handle_pivot_no_index(self, settings: input_schema.NodePivot, var_name: str, input_df: str, agg_func: str):
+        pivot_input = settings.pivot_input
+
+        self._add_code(f'{var_name} = ({input_df}.collect()')
+        self._add_code('    .with_columns(pl.lit(1).alias("__temp_index__"))')
+        self._add_code('    .pivot(')
+        self._add_code(f'        values="{pivot_input.value_col}",')
+        self._add_code(f'        index=["__temp_index__"],')
+        self._add_code(f'        columns="{pivot_input.pivot_column}",')
+        self._add_code(f'        aggregate_function="{agg_func}"')
+        self._add_code("    )")
+        self._add_code('    .drop("__temp_index__")')
+        self._add_code(").lazy()")
+        self._add_code("")
+
+    def _handle_pivot(self, settings: input_schema.NodePivot, var_name: str, input_vars: Dict[str, str]) -> None:
+        """Handle pivot nodes."""
+        input_df = input_vars.get('main', 'df')
+        pivot_input = settings.pivot_input
+        if len(pivot_input.aggregations) > 1:
+            logger.error("Multiple aggregations are not convertable to polars code. "
+                         "Taking the first value")
+        if len(pivot_input.aggregations) > 0:
+            agg_func = pivot_input.aggregations[0]
+        else:
+            agg_func = 'first'
+        if len(settings.pivot_input.index_columns) == 0:
+            self._handle_pivot_no_index(settings, var_name, input_df, agg_func)
+        else:
+            # Generate pivot code
+            self._add_code(f"{var_name} = {input_df}.collect().pivot(")
+            self._add_code(f"    values='{pivot_input.value_col}',")
+            self._add_code(f"    index={pivot_input.index_columns},")
+            self._add_code(f"    columns='{pivot_input.pivot_column}',")
+
+            self._add_code(f"    aggregate_function='{agg_func}'")
+            self._add_code(").lazy()")
+            self._add_code("")
+
+    def _handle_unpivot(self, settings: input_schema.NodeUnpivot, var_name: str, input_vars: Dict[str, str]) -> None:
+        """Handle unpivot nodes."""
+        input_df = input_vars.get('main', 'df')
+        unpivot_input = settings.unpivot_input
+
+        self._add_code(f"{var_name} = {input_df}.unpivot(")
+
+        if unpivot_input.index_columns:
+            self._add_code(f"    index={unpivot_input.index_columns},")
+
+        if unpivot_input.value_columns:
+            self._add_code(f"    on={unpivot_input.value_columns},")
+
+        self._add_code("    variable_name='variable',")
+        self._add_code("    value_name='value'")
+        self._add_code(")")
+        self._add_code("")
+
+    def _handle_union(self, settings: input_schema.NodeUnion, var_name: str, input_vars: Dict[str, str]) -> None:
+        """Handle union nodes."""
+        # Get all input LazyFrame
+        dfs = []
+        if 'main' in input_vars:
+            dfs.append(input_vars['main'])
+        else:
+            # Multiple main inputs
+            for key, df_var in input_vars.items():
+                if key.startswith('main'):
+                    dfs.append(df_var)
+
+        if settings.union_input.mode == 'relaxed':
+            how = 'diagonal_relaxed'
+        else:
+            how = 'diagonal'
+
+        self._add_code(f"{var_name} = pl.concat([")
+        for df in dfs:
+            self._add_code(f"    {df},")
+        self._add_code(f"], how='{how}')")
+        self._add_code("")
+
+    def _handle_sort(self, settings: input_schema.NodeSort, var_name: str, input_vars: Dict[str, str]) -> None:
+        """Handle sort nodes."""
+        input_df = input_vars.get('main', 'df')
+
+        sort_cols = []
+        descending = []
+
+        for sort_input in settings.sort_input:
+            sort_cols.append(f'"{sort_input.column}"')
+            descending.append(sort_input.how == 'desc')
+
+        self._add_code(f"{var_name} = {input_df}.sort([{', '.join(sort_cols)}], descending={descending})")
+        self._add_code("")
+
+    def _handle_sample(self, settings: input_schema.NodeSample, var_name: str, input_vars: Dict[str, str]) -> None:
+        """Handle sample nodes."""
+        input_df = input_vars.get('main', 'df')
+        self._add_code(f"{var_name} = {input_df}.head(n={settings.sample_size})")
+        self._add_code("")
+
+    def _handle_unique(self, settings: input_schema.NodeUnique, var_name: str, input_vars: Dict[str, str]) -> None:
+        """Handle unique/distinct nodes."""
+        input_df = input_vars.get('main', 'df')
+
+        if settings.unique_input.columns:
+            self._add_code(f"{var_name} = {input_df}.unique(subset={settings.unique_input.columns}, keep='{settings.unique_input.strategy}')")
+        else:
+            self._add_code(f"{var_name} = {input_df}.unique(keep='{settings.unique_input.strategy}')")
+        self._add_code("")
+
+    def _handle_text_to_rows(self, settings: input_schema.NodeTextToRows, var_name: str, input_vars: Dict[str, str]) -> None:
+        """Handle text to rows (explode) nodes."""
+        input_df = input_vars.get('main', 'df')
+        text_input = settings.text_to_rows_input
+
+        # First split the column
+        split_expr = f'pl.col("{text_input.column_to_split}").str.split("{text_input.split_fixed_value}")'
+        if text_input.output_column_name and text_input.output_column_name != text_input.column_to_split:
+            split_expr = f'{split_expr}.alias("{text_input.output_column_name}")'
+            explode_col = text_input.output_column_name
+        else:
+            explode_col = text_input.column_to_split
+
+        self._add_code(f"{var_name} = {input_df}.with_columns({split_expr}).explode('{explode_col}')")
+        self._add_code("")
+    # .with_columns(
+    #     (pl.cum_count(record_id_settings.output_column_name)
+    #      .over(record_id_settings.group_by_columns) + record_id_settings.offset - 1)
+    #     .alias(record_id_settings.output_column_name)
+    # )
+    def _handle_record_id(self, settings: input_schema.NodeRecordId, var_name: str, input_vars: Dict[str, str]) -> None:
+        """Handle record ID nodes."""
+        input_df = input_vars.get('main', 'df')
+        record_input = settings.record_id_input
+        if record_input.group_by and record_input.group_by_columns:
+
+            # Row number within groups
+            self._add_code(f"{var_name} = ({input_df}")
+            self._add_code(f"    .with_columns(pl.lit(1).alias('{record_input.output_column_name}'))")
+            self._add_code(f"    .with_columns([")
+            self._add_code(f"    (pl.cum_count('{record_input.output_column_name}').over({record_input.group_by_columns}) + {record_input.offset} - 1)")
+            self._add_code(f"    .alias('{record_input.output_column_name}')")
+            self._add_code("])")
+            self._add_code(f".select(['{record_input.output_column_name}'] + [col for col in {input_df}.columns if col != '{record_input.output_column_name}'])")
+            self._add_code(")")
+        else:
+            # Simple row number
+            self._add_code(f"{var_name} = {input_df}.with_row_count(name='{record_input.output_column_name}', offset={record_input.offset})")
+        self._add_code("")
+
+    def _handle_cross_join(self, settings: input_schema.NodeCrossJoin, var_name: str, input_vars: Dict[str, str]) -> None:
+        """Handle cross join nodes."""
+        left_df = input_vars.get('main', input_vars.get('main_0', 'df_left'))
+        right_df = input_vars.get('right', input_vars.get('main_1', 'df_right'))
+
+        self._add_code(f"{var_name} = {left_df}.join({right_df}, how='cross')")
+        self._add_code("")
+
+    def _handle_output(self, settings: input_schema.NodeOutput, var_name: str, input_vars: Dict[str, str]) -> None:
+        """Handle output nodes."""
+        input_df = input_vars.get('main', 'df')
+        output_settings = settings.output_settings
+
+        if output_settings.file_type == 'csv':
+            self._add_code(f'{input_df}.sink_csv(')
+            self._add_code(f'    "{output_settings.abs_file_path}",')
+            self._add_code(f'    separator="{output_settings.output_csv_table.delimiter}"')
+            self._add_code(')')
+
+        elif output_settings.file_type == 'parquet':
+            self._add_code(f'{input_df}.sink_parquet("{output_settings.abs_file_path}")')
+
+        elif output_settings.file_type == 'excel':
+            self._add_code(f'{input_df}.collect().write_excel(')
+            self._add_code(f'    "{output_settings.abs_file_path}",')
+            self._add_code(f'    worksheet="{output_settings.output_excel_table.sheet_name}"')
+            self._add_code(')')
+
+        self._add_code("")
+
+    def _handle_polars_code(self, settings: input_schema.NodePolarsCode, var_name: str, input_vars: Dict[str, str]) -> None:
+        """Handle custom Polars code nodes."""
+        code = settings.polars_code_input.polars_code.strip()
+        # Determine function parameters based on number of inputs
+        if len(input_vars) == 0:
+            params = ""
+            args = ""
+        elif len(input_vars) == 1:
+            params = "input_df: pl.LazyFrame"
+            input_df = list(input_vars.values())[0]
+            args = input_df
+        else:
+            # Multiple inputs
+            param_list = []
+            arg_list = []
+            i = 1
+            for key in sorted(input_vars.keys()):
+                if key.startswith('main'):
+                    param_list.append(f"input_df_{i}: pl.LazyFrame")
+                    arg_list.append(input_vars[key])
+                    i += 1
+            params = ", ".join(param_list)
+            args = ", ".join(arg_list)
+
+        # Check if the code is just an expression (no assignment)
+        is_expression = '=' not in code or code.count('=') == 1 and '==' in code
+
+        # Wrap the code in a function
+        self._add_code(f"# Custom Polars code")
+        self._add_code(f"def _polars_code_{var_name.replace('df_', '')}({params}):")
+
+        # Handle the code based on its structure
+        if is_expression:
+            # It's just an expression, return it directly
+            self._add_code(f"    return {code}")
+        else:
+            # It contains assignments
+            for line in code.split('\n'):
+                if line.strip():
+                    self._add_code(f"    {line}")
+
+            # If no explicit return, try to detect what to return
+            if 'return' not in code:
+                # Try to find the last assignment
+                lines = [l.strip() for l in code.split('\n') if l.strip() and '=' in l]
+                if lines:
+                    last_assignment = lines[-1]
+                    if '=' in last_assignment:
+                        output_var = last_assignment.split('=')[0].strip()
+                        self._add_code(f"    return {output_var}")
+
+        self._add_code("")
+
+        # Call the function
+        self._add_code(f"{var_name} = _polars_code_{var_name.replace('df_', '')}({args})")
+        self._add_code("")
+
+    # Helper methods
+
+    def _add_code(self, line: str) -> None:
+        """Add a line of code."""
+        self.code_lines.append(line)
+
+    def _add_comment(self, comment: str) -> None:
+        """Add a comment line."""
+        self.code_lines.append(comment)
+
+    def _parse_filter_expression(self, expr: str) -> str:
+        """Parse Flowfile filter expression to Polars expression."""
+        # This is a simplified parser - you'd need more sophisticated parsing
+        # Handle patterns like [column]>value or [column]="value"
+
+        import re
+
+        # Pattern: [column_name]operator"value" or [column_name]operatorvalue
+        pattern = r'\[([^\]]+)\]([><=!]+)"?([^"]*)"?'
+
+        def replace_expr(match):
+            col, op, val = match.groups()
+
+            # Map operators
+            op_map = {
+                '=': '==',
+                '!=': '!=',
+                '>': '>',
+                '<': '<',
+                '>=': '>=',
+                '<=': '<='
+            }
+
+            polars_op = op_map.get(op, op)
+
+            # Check if value is numeric
+            try:
+                float(val)
+                return f'pl.col("{col}") {polars_op} {val}'
+            except ValueError:
+                return f'pl.col("{col}") {polars_op} "{val}"'
+
+        return re.sub(pattern, replace_expr, expr)
+
+    def _create_basic_filter_expr(self, basic: transform_schema.BasicFilter) -> str:
+        """Create Polars expression from basic filter."""
+        col = f'pl.col("{basic.field}")'
+
+        if basic.filter_type == 'equals':
+            return f'{col} == "{basic.filter_value}"'
+        elif basic.filter_type == 'not_equals':
+            return f'{col} != "{basic.filter_value}"'
+        elif basic.filter_type == 'greater':
+            return f'{col} > {basic.filter_value}'
+        elif basic.filter_type == 'less':
+            return f'{col} < {basic.filter_value}'
+        elif basic.filter_type == 'in':
+            values = basic.filter_value.split(',')
+            return f"pl.col('{col}').is_in({values})"
+        return col
+
+    def _get_polars_dtype(self, dtype_str: str) -> str:
+        """Convert Flowfile dtype string to Polars dtype."""
+        dtype_map = {
+            'String': 'pl.Utf8',
+            'Integer': 'pl.Int64',
+            'Double': 'pl.Float64',
+            'Boolean': 'pl.Boolean',
+            'Date': 'pl.Date',
+            'Datetime': 'pl.Datetime',
+            'Float32': 'pl.Float32',
+            'Float64': 'pl.Float64',
+            'Int32': 'pl.Int32',
+            'Int64': 'pl.Int64',
+            'Utf8': 'pl.Utf8',
+        }
+        return dtype_map.get(dtype_str, 'pl.Utf8')
+
+    def _get_agg_function(self, agg: str) -> str:
+        """Get Polars aggregation function name."""
+        agg_map = {
+            'avg': 'mean',
+            'average': 'mean',
+            'concat': 'str.concat',
+        }
+        return agg_map.get(agg, agg)
+
+    def _sql_to_polars_expr(self, sql_expr: str) -> str:
+        """Convert SQL-like expression to Polars expression."""
+        # This is a very simplified converter
+        # In practice, you'd want a proper SQL parser
+
+        # Replace column references
+        import re
+
+        # Pattern for column names (simplified)
+        col_pattern = r'\b([a-zA-Z_][a-zA-Z0-9_]*)\b'
+
+        def replace_col(match):
+            col_name = match.group(1)
+            # Skip SQL keywords
+            keywords = {'CASE', 'WHEN', 'THEN', 'ELSE', 'END', 'AND', 'OR', 'NOT', 'IN', 'AS'}
+            if col_name.upper() in keywords:
+                return col_name
+            return f'pl.col("{col_name}")'
+
+        result = re.sub(col_pattern, replace_col, sql_expr)
+
+        # Handle CASE WHEN
+        if 'CASE' in result:
+            # This would need proper parsing
+            result = "pl.when(...).then(...).otherwise(...)"
+
+        return result
+
+    def add_return_code(self, lines: List[str]) -> None:
+        if self.output_nodes:
+            # Return marked output nodes
+            if len(self.output_nodes) == 1:
+                # Single output
+                _, var_name = self.output_nodes[0]
+                lines.append(f"    return {var_name}")
+            else:
+                # Multiple outputs - return as dictionary
+                lines.append("    return {")
+                for node_id, var_name in self.output_nodes:
+                    lines.append(f'        "node_{node_id}": {var_name},')
+                lines.append("    }")
+        elif self.last_node_var:
+            lines.append(f"    return {self.last_node_var}")
+        else:
+            lines.append("    return None")
+
+    def _build_final_code(self) -> str:
+        """Build the final Python code."""
+        lines = []
+
+        # Add imports
+        lines.extend(sorted(self.imports))
+        lines.append("")
+        lines.append("")
+
+        # Add main function
+        lines.append("def run_etl_pipeline():")
+        lines.append('    """')
+        lines.append(f'    ETL Pipeline: {self.flow_graph.__name__}')
+        lines.append('    Generated from Flowfile')
+        lines.append('    """')
+        lines.append("    ")
+
+        # Add the generated code
+        for line in self.code_lines:
+            if line:
+                lines.append(f"    {line}")
+            else:
+                lines.append("")
+        # Add main block
+        lines.append("")
+        self.add_return_code(lines)
+        lines.append("")
+        lines.append("")
+        lines.append('if __name__ == "__main__":')
+        lines.append("    pipeline_output = run_etl_pipeline()")
+
+        return "\n".join(lines)
+
+
+# Example usage function
+def export_flow_to_polars(flow_graph: FlowGraph) -> str:
+    """
+    Export a FlowGraph to standalone Polars code.
+
+    Args:
+        flow_graph: The FlowGraph instance to convert
+
+    Returns:
+        str: Python code that can be executed standalone
+    """
+    converter = FlowGraphToPolarsConverter(flow_graph)
+    return converter.convert()