Ma77Ball commented on code in PR #6663: URL: https://github.com/apache/texera/pull/6663#discussion_r3627121495
########## common/workflow-operator/src/test/scala/org/apache/texera/amber/operator/huggingFace/codegen/PythonCodegenBaseSpec.scala: ########## @@ -0,0 +1,131 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.texera.amber.operator.huggingFace.codegen + +import org.apache.texera.amber.pybuilder.PyStringTypes.EncodableString +import org.scalatest.flatspec.AnyFlatSpec +import org.scalatest.matchers.should.Matchers + +class PythonCodegenBaseSpec extends AnyFlatSpec with Matchers { + + // A stand-in TaskCodegen whose payload/parse snippets are distinctive + // sentinels. Because payloadPython/parsePython return plain Python source + // fragments (not EncodableString user data), the base template must splice + // them in verbatim, letting the spec prove the splice happened and that the + // fragment is codegen-driven rather than hardcoded in the base. + private object StubCodegen extends TaskCodegen { + override def task: String = "text-generation" + override def payloadPython(ctx: CodegenContext): String = + " payload = STUB_PAYLOAD_zX7q42" + override def parsePython(ctx: CodegenContext): String = + " return STUB_PARSE_zX7q42" + } + + private def makeCtx( + hfApiToken: EncodableString = "token", + modelId: EncodableString = "Qwen/Qwen2.5-72B-Instruct", + promptColumn: EncodableString = "prompt", + resultColumn: EncodableString = "hf_response", + task: EncodableString = "text-generation", + systemPrompt: EncodableString = "You are a helpful assistant.", + safeMaxTokens: Int = 256, + safeTemp: Double = 0.7 + ): CodegenContext = + CodegenContext( + hfApiToken = hfApiToken, + modelId = modelId, + promptColumn = promptColumn, + resultColumn = resultColumn, + task = task, + systemPrompt = systemPrompt, + safeMaxTokens = safeMaxTokens, + safeTemp = safeTemp + ) + + "PythonCodegenBase.render" should "emit the ProcessTableOperator skeleton and shared helpers" in { + val out = PythonCodegenBase.render(makeCtx(), StubCodegen) + out should include("class ProcessTableOperator(UDFTableOperator):") + out should include("def open(self):") + out should include("def process_table(") + out should include("def _parse_response(") + out should include("def _resolve_providers(") + out should include("def _post_with_fallback(") + out should include("PROVIDER_COST_PRIORITY") + out should include("CHAT_ROUTES") + out should include("MAX_REMOTE_FETCH_BYTES") + } + + it should "splice the per-task codegen's payload and parse snippets into the template" in { + val out = PythonCodegenBase.render(makeCtx(), StubCodegen) + out should include("STUB_PAYLOAD_zX7q42") + out should include("STUB_PARSE_zX7q42") + } + + it should "delegate payload/parse to the codegen rather than hardcoding a task's output" in { + // The same context routed through two different codegens must yield the + // spliced fragments of whichever codegen is passed, proving the base is a + // pure host for the codegen's snippets. + val stub = PythonCodegenBase.render(makeCtx(), StubCodegen) + val real = PythonCodegenBase.render(makeCtx(), TextGenCodegen) + stub should include("STUB_PAYLOAD_zX7q42") + real should not include "STUB_PAYLOAD_zX7q42" + // TextGenCodegen's real payload/parse markers appear only in the real run. + real should include("messages") + real should include("max_tokens") + real should include("choices") + } + + it should "interpolate the numeric context fields as raw Python literals" in { + val out = PythonCodegenBase.render(makeCtx(safeMaxTokens = 512, safeTemp = 0.9), StubCodegen) + out should include("self.MAX_NEW_TOKENS = 512") + out should include("self.TEMPERATURE = 0.9") + } + + it should "assign user-provided strings via runtime base64 decode expressions, not raw literals" in { + val out = PythonCodegenBase.render(makeCtx(), StubCodegen) + // Every user-supplied string field is set through the safe decode helper. + out should include("self.HF_API_TOKEN = self.decode_python_template(") + out should include("self.MODEL_ID = self.decode_python_template(") + out should include("self.PROMPT_COLUMN = self.decode_python_template(") + out should include("self.SYSTEM_PROMPT = self.decode_python_template(") + } Review Comment: Good catch. I extended the assertions to cover every EncodableString field `open()` assigns via `decode_python_template` (RESULT_COLUMN, TASK, and the image/audio/context/labels/sentences fields), not just the text-gen subset. ########## common/workflow-operator/src/test/scala/org/apache/texera/amber/operator/huggingFace/codegen/PythonCodegenBaseSpec.scala: ########## @@ -0,0 +1,131 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.texera.amber.operator.huggingFace.codegen + +import org.apache.texera.amber.pybuilder.PyStringTypes.EncodableString +import org.scalatest.flatspec.AnyFlatSpec +import org.scalatest.matchers.should.Matchers + +class PythonCodegenBaseSpec extends AnyFlatSpec with Matchers { + + // A stand-in TaskCodegen whose payload/parse snippets are distinctive + // sentinels. Because payloadPython/parsePython return plain Python source + // fragments (not EncodableString user data), the base template must splice + // them in verbatim, letting the spec prove the splice happened and that the + // fragment is codegen-driven rather than hardcoded in the base. + private object StubCodegen extends TaskCodegen { + override def task: String = "text-generation" + override def payloadPython(ctx: CodegenContext): String = + " payload = STUB_PAYLOAD_zX7q42" + override def parsePython(ctx: CodegenContext): String = + " return STUB_PARSE_zX7q42" + } + + private def makeCtx( + hfApiToken: EncodableString = "token", + modelId: EncodableString = "Qwen/Qwen2.5-72B-Instruct", + promptColumn: EncodableString = "prompt", + resultColumn: EncodableString = "hf_response", + task: EncodableString = "text-generation", + systemPrompt: EncodableString = "You are a helpful assistant.", + safeMaxTokens: Int = 256, + safeTemp: Double = 0.7 + ): CodegenContext = + CodegenContext( + hfApiToken = hfApiToken, + modelId = modelId, + promptColumn = promptColumn, + resultColumn = resultColumn, + task = task, + systemPrompt = systemPrompt, + safeMaxTokens = safeMaxTokens, + safeTemp = safeTemp + ) + + "PythonCodegenBase.render" should "emit the ProcessTableOperator skeleton and shared helpers" in { + val out = PythonCodegenBase.render(makeCtx(), StubCodegen) + out should include("class ProcessTableOperator(UDFTableOperator):") + out should include("def open(self):") + out should include("def process_table(") + out should include("def _parse_response(") + out should include("def _resolve_providers(") + out should include("def _post_with_fallback(") + out should include("PROVIDER_COST_PRIORITY") + out should include("CHAT_ROUTES") + out should include("MAX_REMOTE_FETCH_BYTES") + } + + it should "splice the per-task codegen's payload and parse snippets into the template" in { + val out = PythonCodegenBase.render(makeCtx(), StubCodegen) + out should include("STUB_PAYLOAD_zX7q42") + out should include("STUB_PARSE_zX7q42") + } + + it should "delegate payload/parse to the codegen rather than hardcoding a task's output" in { + // The same context routed through two different codegens must yield the + // spliced fragments of whichever codegen is passed, proving the base is a + // pure host for the codegen's snippets. + val stub = PythonCodegenBase.render(makeCtx(), StubCodegen) + val real = PythonCodegenBase.render(makeCtx(), TextGenCodegen) + stub should include("STUB_PAYLOAD_zX7q42") + real should not include "STUB_PAYLOAD_zX7q42" + // TextGenCodegen's real payload/parse markers appear only in the real run. + real should include("messages") + real should include("max_tokens") + real should include("choices") Review Comment: Right, `messages` shows up in several base-template helper branches. I replaced it with TextGenCodegen's unique system-role message, which I confirmed does not appear in the base template. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
