以下是我的textractUtils.js代码 –

const _ = require("lodash");const aws = require("aws-sdk");const config = require("./config");aws.config.update({  accessKeyId: config.awsAccesskeyID,  secretAccessKey: config.awsSecretAccessKey,  region: config.awsRegion});const textract = new aws.Textract();const getText = (result, blocksMap) => {  let text = "";  if (_.has(result, "Relationships")) {    result.Relationships.forEach(relationship => {      if (relationship.Type === "CHILD") {        relationship.Ids.forEach(childId => {          const word = blocksMap[childId];          if (word.BlockType === "WORD") {            text += `${word.Text} `;          }          if (word.BlockType === "SELECTION_ELEMENT") {            if (word.SelectionStatus === "SELECTED") {              text += `X `;            }          }        });      }    });  }  return text.trim();};const findValueBlock = (keyBlock, valueMap) => {  let valueBlock;  keyBlock.Relationships.forEach(relationship => {    if (relationship.Type === "VALUE") {      // eslint-disable-next-line array-callback-return      relationship.Ids.every(valueId => {        if (_.has(valueMap, valueId)) {          valueBlock = valueMap[valueId];          return false;        }      });    }  });  return valueBlock;};const getKeyValueRelationship = (keyMap, valueMap, blockMap) => {  const keyValues = {};  const keyMapValues = _.values(keyMap);  keyMapValues.forEach(keyMapValue => {    const valueBlock = findValueBlock(keyMapValue, valueMap);    const key = getText(keyMapValue, blockMap);    const value = getText(valueBlock, blockMap);    keyValues[key] = value;  });  return keyValues;};const getKeyValueMap = blocks => {  const keyMap = {};  const valueMap = {};  const blockMap = {};  let blockId;  blocks.forEach(block => {    blockId = block.Id;    blockMap[blockId] = block;    if (block.BlockType === "KEY_VALUE_SET") {      if (_.includes(block.EntityTypes, "KEY")) {        keyMap[blockId] = block;      } else {        valueMap[blockId] = block;      }    }  });  return { keyMap, valueMap, blockMap };};module.exports = async buffer => {  const params = {    Document: {      /* required */      Bytes: buffer    },    FeatureTypes: ["FORMS"]  };  const request = textract.analyzeDocument(params);  const data = await request.promise();  if (data && data.Blocks) {    const { keyMap, valueMap, blockMap } = getKeyValueMap(data.Blocks);    const keyValues = getKeyValueRelationship(keyMap, valueMap, blockMap);    return keyValues;  }  // 如果没有找到块，则返回undefined  return undefined;};

它在处理图片时运行正常，但处理pdf时（无论是单页还是多页）都不行。以下是我导入pdf时遇到的错误 –

(node:2001) UnhandledPromiseRejectionWarning: UnsupportedDocumentException: Request has unsupported document format    at Request.extractError (/home/<user>/textract-lab/node_modules/aws-sdk/lib/protocol/json.js:51:27)    at Request.callListeners (/home/<user>/textract-lab/node_modules/aws-sdk/lib/sequential_executor.js:106:20)    at Request.emit (/home/<user>/textract-lab/node_modules/aws-sdk/lib/sequential_executor.js:78:10)    at Request.emit (/home/<user>/textract-lab/node_modules/aws-sdk/lib/request.js:683:14)    at Request.transition (/home/<user>/textract-lab/node_modules/aws-sdk/lib/request.js:22:10)    at AcceptorStateMachine.runTo (/home/<user>/textract-lab/node_modules/aws-sdk/lib/state_machine.js:14:12)    at /home/<user>/textract-lab/node_modules/aws-sdk/lib/state_machine.js:26:10    at Request.<anonymous> (/home/<user>/textract-lab/node_modules/aws-sdk/lib/request.js:38:9)    at Request.<anonymous> (/home/<user>/textract-lab/node_modules/aws-sdk/lib/request.js:685:12)    at Request.callListeners (/home/<user>/textract-lab/node_modules/aws-sdk/lib/sequential_executor.js:116:18)(node:2001) UnhandledPromiseRejectionWarning: Unhandled promise rejection. This error originated either by throwing inside of an async function without a catch block, or by rejecting a promise which was not handled with .catch(). (rejection id: 1)(node:2001) [DEP0018] DeprecationWarning: Unhandled promise rejections are deprecated. In the future, promise rejections that are not handled will terminate the Node.js process with a non-zero exit code.

我尝试过处理不含文本的图片、含文本的图片、含表格的图片、单页pdf和多页pdf。我还有一个概念上的疑问，如果我已经导入了aws-sdk，为什么我还需要为pdf编写代码，因为aws-sdk的textract应该会处理pdf、png、jpeg和jpg格式的图片？我需要对textractUtils.js做哪些修改才能处理pdf文件？

回答：

AnalyzeDocument API操作只支持PNG或JPEG格式的图片。根据Textract的文档：

Amazon Textract同步操作（DetectDocumentText和AnalyzeDocument）支持PNG和JPEG图像格式。异步操作（StartDocumentTextDetection、StartDocumentAnalysis）也支持PDF文件格式。

您应该使用异步操作来处理您的PDF文档。否则，一个解决方法是在您的代码中将PDF文档转换为图片，然后使用这些图片的同步API操作来处理文档。