|
1 | 1 | from importlib import import_module |
2 | 2 | import json |
3 | | -import traceback |
4 | 3 |
|
5 | | -from aiida import orm |
6 | | -from aiida_pythonjob.data.serializer import general_serializer |
7 | | -from aiida_workgraph import WorkGraph, task |
8 | | -from aiida_workgraph.socket import TaskSocketNamespace |
9 | 4 |
|
10 | 5 | from python_workflow_definition.shared import ( |
11 | 6 | convert_nodes_list_to_dict, |
|
18 | 13 | ) |
19 | 14 |
|
20 | 15 |
|
21 | | -def load_workflow_json(file_name: str) -> WorkGraph: |
| 16 | +# -> WorkGraph |
| 17 | +def load_workflow_json(file_name: str): |
| 18 | + """ |
| 19 | + Load a workflow JSON file and convert it to a WorkGraph. |
| 20 | +
|
| 21 | + Args: |
| 22 | + file_name: Path to the JSON workflow file |
| 23 | +
|
| 24 | + Returns: |
| 25 | + A populated WorkGraph object |
| 26 | + """ |
22 | 27 | with open(file_name) as f: |
23 | 28 | data = json.load(f) |
24 | 29 |
|
| 30 | + from aiida_workgraph import WorkGraph |
25 | 31 | wg = WorkGraph() |
26 | 32 | task_name_mapping = {} |
27 | 33 |
|
28 | | - for id, identifier in convert_nodes_list_to_dict( |
29 | | - nodes_list=data[NODES_LABEL] |
30 | | - ).items(): |
| 34 | + # Create all tasks first |
| 35 | + for task_id, identifier in convert_nodes_list_to_dict(data[NODES_LABEL]).items(): |
31 | 36 | if isinstance(identifier, str) and "." in identifier: |
32 | | - p, m = identifier.rsplit(".", 1) |
33 | | - mod = import_module(p) |
34 | | - func = getattr(mod, m) |
| 37 | + # Import the function dynamically |
| 38 | + package_path, function_name = identifier.rsplit(".", 1) |
| 39 | + module = import_module(package_path) |
| 40 | + func = getattr(module, function_name) |
| 41 | + |
| 42 | + # Add task and prepare for linking |
35 | 43 | wg.add_task(func) |
36 | | - # Remove the default result output, because we will add the outputs later from the data in the link |
37 | | - del wg.tasks[-1].outputs["result"] |
38 | | - task_name_mapping[id] = wg.tasks[-1] |
39 | | - else: |
40 | | - # data task |
41 | | - data_node = general_serializer(identifier) |
42 | | - task_name_mapping[id] = data_node |
43 | | - |
44 | | - # add links |
| 44 | + current_task = wg.tasks[-1] |
| 45 | + |
| 46 | + # Remove default output as we'll add custom outputs later |
| 47 | + del current_task.outputs["result"] |
| 48 | + task_name_mapping[task_id] = current_task |
| 49 | + |
| 50 | + # Add all connections between tasks |
45 | 51 | for link in data[EDGES_LABEL]: |
46 | | - to_task = task_name_mapping[str(link[TARGET_LABEL])] |
47 | | - # if the input is not exit, it means we pass the data into to the kwargs |
48 | | - # in this case, we add the input socket |
49 | | - if link[TARGET_PORT_LABEL] not in to_task.inputs: |
50 | | - to_socket = to_task.add_input("workgraph.any", name=link[TARGET_PORT_LABEL]) |
51 | | - else: |
52 | | - to_socket = to_task.inputs[link[TARGET_PORT_LABEL]] |
53 | | - from_task = task_name_mapping[str(link[SOURCE_LABEL])] |
54 | | - if isinstance(from_task, orm.Data): |
55 | | - to_socket.value = from_task |
56 | | - else: |
57 | | - try: |
58 | | - if link[SOURCE_PORT_LABEL] is None: |
59 | | - link[SOURCE_PORT_LABEL] = "result" |
60 | | - # because we are not define the outputs explicitly during the pythonjob creation |
61 | | - # we add it here, and assume the output exit |
62 | | - if link[SOURCE_PORT_LABEL] not in from_task.outputs: |
63 | | - # if str(link["sourcePort"]) not in from_task.outputs: |
64 | | - from_socket = from_task.add_output( |
65 | | - "workgraph.any", |
66 | | - name=link[SOURCE_PORT_LABEL], |
67 | | - # name=str(link["sourcePort"]), |
68 | | - metadata={"is_function_output": True}, |
69 | | - ) |
70 | | - else: |
71 | | - from_socket = from_task.outputs[link[SOURCE_PORT_LABEL]] |
72 | | - |
73 | | - wg.add_link(from_socket, to_socket) |
74 | | - except Exception as e: |
75 | | - traceback.print_exc() |
76 | | - print("Failed to link", link, "with error:", e) |
| 52 | + source_id = link[SOURCE_LABEL] |
| 53 | + target_id = link[TARGET_LABEL] |
| 54 | + source_port = link[SOURCE_PORT_LABEL] |
| 55 | + target_port = link[TARGET_PORT_LABEL] |
| 56 | + |
| 57 | + # Handle task-to-task connections |
| 58 | + if source_id is not None and target_id is not None: |
| 59 | + from_task = task_name_mapping[str(source_id)] |
| 60 | + to_task = task_name_mapping[str(target_id)] |
| 61 | + |
| 62 | + # Create output socket on source task |
| 63 | + from_socket = from_task.add_output("workgraph.any", name=source_port, metadata={"is_function_output": True}) |
| 64 | + |
| 65 | + # Create or get input socket on target task |
| 66 | + if target_port not in to_task.inputs: |
| 67 | + to_socket = to_task.add_input("workgraph.any", name=target_port) |
| 68 | + else: |
| 69 | + to_socket = to_task.inputs[target_port] |
| 70 | + |
| 71 | + # Connect the tasks |
| 72 | + wg.add_link(from_socket, to_socket) |
| 73 | + |
| 74 | + # Handle dangling outputs (no target) |
| 75 | + elif source_id is not None and target_id is None: |
| 76 | + from_task = task_name_mapping[str(source_id)] |
| 77 | + from_task.add_output("workgraph.any", name=source_port, metadata={"is_function_output": True}) |
| 78 | + |
77 | 79 | return wg |
78 | 80 |
|
79 | 81 |
|
80 | | -def write_workflow_json(wg: WorkGraph, file_name: str) -> dict: |
| 82 | +def write_workflow_json(wg: "WorkGraph", file_name: str) -> dict: |
| 83 | + """ |
| 84 | + Write a WorkGraph to a JSON file. |
| 85 | +
|
| 86 | + Args: |
| 87 | + wg: WorkGraph object to serialize |
| 88 | + file_name: Path where the JSON file will be written |
| 89 | +
|
| 90 | + Returns: |
| 91 | + Dictionary representation of the serialized workflow |
| 92 | + """ |
81 | 93 | data = {NODES_LABEL: [], EDGES_LABEL: []} |
82 | 94 | node_name_mapping = {} |
83 | | - data_node_name_mapping = {} |
84 | | - i = 0 |
85 | | - for node in wg.tasks: |
86 | | - executor = node.get_executor() |
| 95 | + |
| 96 | + # Add all process nodes first |
| 97 | + for i, node in enumerate(wg.tasks): |
| 98 | + # Store node index for later reference |
87 | 99 | node_name_mapping[node.name] = i |
88 | 100 |
|
89 | | - callable_name = executor["callable_name"] |
90 | | - callable_name = f"{executor['module_path']}.{callable_name}" |
| 101 | + # Get executor info and build full callable name |
| 102 | + executor = node.get_executor() |
| 103 | + callable_name = f"{executor['module_path']}.{executor['callable_name']}" |
| 104 | + |
| 105 | + # Add node to data structure |
91 | 106 | data[NODES_LABEL].append({"id": i, "function": callable_name}) |
92 | | - i += 1 |
93 | 107 |
|
| 108 | + # Create edges from WorkGraph links |
94 | 109 | for link in wg.links: |
95 | 110 | link_data = link.to_dict() |
96 | | - # if the from socket is the default result, we set it to None |
97 | | - if link_data["from_socket"] == "result": |
98 | | - link_data["from_socket"] = None |
99 | | - link_data[TARGET_LABEL] = node_name_mapping[link_data.pop("to_node")] |
100 | | - link_data[TARGET_PORT_LABEL] = link_data.pop("to_socket") |
101 | | - link_data[SOURCE_LABEL] = node_name_mapping[link_data.pop("from_node")] |
102 | | - link_data[SOURCE_PORT_LABEL] = link_data.pop("from_socket") |
103 | | - data[EDGES_LABEL].append(link_data) |
104 | | - |
105 | | - for node in wg.tasks: |
106 | | - for input in node.inputs: |
107 | | - # assume namespace is not used as input |
108 | | - if isinstance(input, TaskSocketNamespace): |
109 | | - continue |
110 | | - if isinstance(input.value, orm.Data): |
111 | | - if input.value.uuid not in data_node_name_mapping: |
112 | | - if isinstance(input.value, orm.List): |
113 | | - raw_value = input.value.get_list() |
114 | | - elif isinstance(input.value, orm.Dict): |
115 | | - raw_value = input.value.get_dict() |
116 | | - # unknow reason, there is a key "node_type" in the dict |
117 | | - raw_value.pop("node_type", None) |
118 | | - else: |
119 | | - raw_value = input.value.value |
120 | | - data[NODES_LABEL].append({"id": i, "value": raw_value}) |
121 | | - input_node_name = i |
122 | | - data_node_name_mapping[input.value.uuid] = input_node_name |
123 | | - i += 1 |
124 | | - else: |
125 | | - input_node_name = data_node_name_mapping[input.value.uuid] |
126 | | - data[EDGES_LABEL].append( |
127 | | - { |
128 | | - TARGET_LABEL: node_name_mapping[node.name], |
129 | | - TARGET_PORT_LABEL: input._name, |
130 | | - SOURCE_LABEL: input_node_name, |
131 | | - SOURCE_PORT_LABEL: None, |
132 | | - } |
133 | | - ) |
| 111 | + |
| 112 | + # Handle default result case |
| 113 | + from_socket = link_data.pop("from_socket") |
| 114 | + if from_socket == "result": |
| 115 | + from_socket = None |
| 116 | + |
| 117 | + # Convert to expected format |
| 118 | + edge = { |
| 119 | + SOURCE_LABEL: node_name_mapping[link_data.pop("from_node")], |
| 120 | + SOURCE_PORT_LABEL: from_socket, |
| 121 | + TARGET_LABEL: node_name_mapping[link_data.pop("to_node")], |
| 122 | + TARGET_PORT_LABEL: link_data.pop("to_socket"), |
| 123 | + } |
| 124 | + |
| 125 | + data[EDGES_LABEL].append(edge) |
| 126 | + |
| 127 | + # Define sockets to ignore when adding workflow I/O connections |
| 128 | + wg_input_sockets = { |
| 129 | + "_wait", |
| 130 | + "metadata", |
| 131 | + "function_data", |
| 132 | + "process_label", |
| 133 | + "function_inputs", |
| 134 | + "deserializers", |
| 135 | + "serializers", |
| 136 | + } |
| 137 | + wg_output_sockets = {"_wait", "_outputs", "exit_code"} |
| 138 | + |
| 139 | + # Handle first node's inputs (external inputs to workflow) |
| 140 | + first_node = wg.tasks[0] |
| 141 | + first_node_id = node_name_mapping[first_node.name] |
| 142 | + |
| 143 | + for input_name in first_node.inputs._sockets: |
| 144 | + if input_name not in wg_input_sockets: |
| 145 | + data[EDGES_LABEL].append( |
| 146 | + { |
| 147 | + SOURCE_LABEL: None, |
| 148 | + SOURCE_PORT_LABEL: None, |
| 149 | + TARGET_LABEL: first_node_id, |
| 150 | + TARGET_PORT_LABEL: input_name, |
| 151 | + } |
| 152 | + ) |
| 153 | + |
| 154 | + # Handle last node's outputs (workflow outputs) |
| 155 | + last_node = wg.tasks[-1] |
| 156 | + last_node_id = node_name_mapping[last_node.name] |
| 157 | + |
| 158 | + for output_name in last_node.outputs._sockets: |
| 159 | + if output_name not in wg_output_sockets: |
| 160 | + data[EDGES_LABEL].append( |
| 161 | + { |
| 162 | + SOURCE_LABEL: last_node_id, |
| 163 | + SOURCE_PORT_LABEL: output_name, |
| 164 | + TARGET_LABEL: None, |
| 165 | + TARGET_PORT_LABEL: None, |
| 166 | + } |
| 167 | + ) |
| 168 | + |
| 169 | + # Write the data to file |
| 170 | + import ipdb; ipdb.set_trace() |
| 171 | + |
134 | 172 | with open(file_name, "w") as f: |
135 | | - # json.dump({"nodes": data[], "edges": edges_new_lst}, f) |
136 | 173 | json.dump(data, f, indent=2) |
137 | 174 |
|
138 | 175 | return data |
0 commit comments