resume-parser-ts/dist/index.cjs.map

A TypeScript library for parsing resumes from PDF files

{"version":3,"sources":["../src/index.ts","../src/lib/parse-resume-from-pdf/read-pdf.ts","../src/lib/parse-resume-from-pdf/extract-resume-from-sections/lib/bullet-points.ts","../src/lib/parse-resume-from-pdf/group-text-items-into-lines.ts","../src/lib/parse-resume-from-pdf/extract-resume-from-sections/lib/common-features.ts","../src/lib/parse-resume-from-pdf/group-lines-into-sections.ts","../src/lib/parse-resume-from-pdf/extract-resume-from-sections/lib/get-section-lines.ts","../src/lib/parse-resume-from-pdf/extract-resume-from-sections/lib/feature-scoring-system.ts","../src/lib/parse-resume-from-pdf/extract-resume-from-sections/extract-profile.ts","../src/lib/parse-resume-from-pdf/extract-resume-from-sections/lib/subsections.ts","../src/lib/parse-resume-from-pdf/extract-resume-from-sections/extract-education.ts","../src/lib/parse-resume-from-pdf/extract-resume-from-sections/extract-work-experience.ts","../src/lib/parse-resume-from-pdf/extract-resume-from-sections/extract-project.ts","../src/lib/deep-clone.ts","../src/lib/redux/resumeSlice.ts","../src/lib/parse-resume-from-pdf/extract-resume-from-sections/extract-skills.ts","../src/lib/parse-resume-from-pdf/extract-resume-from-sections/index.ts","../src/lib/parse-resume-from-pdf/index.ts"],"sourcesContent":["export { parseResumeFromPdf } from './lib/parse-resume-from-pdf/index.js';\nexport type { \n  Resume, \n  ResumeProfile, \n  ResumeWorkExperience, \n  ResumeEducation, \n  ResumeProject, \n  ResumeSkills,\n  ResumeCustom,\n  FeaturedSkill\n} from './lib/redux/types.js';\nexport type { \n  TextItem, \n  TextItems, \n  Line, \n  Lines \n} from './lib/parse-resume-from-pdf/types.js';\n","// Getting pdfjs to work is tricky. The following 3 lines would make it work\r\n// https://stackoverflow.com/a/63486898/7699841\r\nimport * as pdfjs from \"pdfjs-dist\";\r\n// @ts-ignore\r\nimport pdfjsWorker from \"pdfjs-dist/build/pdf.worker.entry\";\r\npdfjs.GlobalWorkerOptions.workerSrc = pdfjsWorker;\r\n\r\nimport type { TextItem as PdfjsTextItem } from \"pdfjs-dist/types/src/display/api\";\r\nimport type { TextItem, TextItems } from \"../../lib/parse-resume-from-pdf/types\";\r\n\r\n/**\r\n * Step 1: Read pdf and output textItems by concatenating results from each page.\r\n *\r\n * To make processing easier, it returns a new TextItem type, which removes unused\r\n * attributes (dir, transform), adds x and y positions, and replaces loaded font\r\n * name with original font name.\r\n *\r\n * @example\r\n * const onFileChange = async (e) => {\r\n *     const fileUrl = URL.createObjectURL(e.target.files[0]);\r\n *     const textItems = await readPdf(fileUrl);\r\n * }\r\n */\r\nexport const readPdf = async (fileUrl: string): Promise<TextItems> => {\r\n  const pdfFile = await pdfjs.getDocument(fileUrl).promise;\r\n  let textItems: TextItems = [];\r\n\r\n  for (let i = 1; i <= pdfFile.numPages; i++) {\r\n    // Parse each page into text content\r\n    const page = await pdfFile.getPage(i);\r\n    const textContent = await page.getTextContent();\r\n\r\n    // Wait for font data to be loaded\r\n    await page.getOperatorList();\r\n    const commonObjs = page.commonObjs;\r\n\r\n    // Convert Pdfjs TextItem type to new TextItem type\r\n    const pageTextItems = textContent.items.map((item) => {\r\n      const {\r\n        str: text,\r\n        dir, // Remove text direction\r\n        transform,\r\n        fontName: pdfFontName,\r\n        ...otherProps\r\n      } = item as PdfjsTextItem;\r\n\r\n      // Extract x, y position of text item from transform.\r\n      // As a side note, origin (0, 0) is bottom left.\r\n      // Reference: https://github.com/mozilla/pdf.js/issues/5643#issuecomment-496648719\r\n      const x = transform[4];\r\n      const y = transform[5];\r\n\r\n      // Use commonObjs to convert font name to original name (e.g. \"GVDLYI+Arial-BoldMT\")\r\n      // since non system font name by default is a loaded name, e.g. \"g_d8_f1\"\r\n      // Reference: https://github.com/mozilla/pdf.js/pull/15659\r\n      const fontObj = commonObjs.get(pdfFontName);\r\n      const fontName = fontObj.name;\r\n\r\n      // pdfjs reads a \"-\" as \"-­‐\" in the resume example. This is to revert it.\r\n      // Note \"-­‐\" is \"-&#x00AD;‐\" with a soft hyphen in between. It is not the same as \"--\"\r\n      const newText = text.replace(/-­‐/g, \"-\");\r\n\r\n      const newItem = {\r\n        ...otherProps,\r\n        fontName,\r\n        text: newText,\r\n        x,\r\n        y,\r\n      };\r\n      return newItem;\r\n    });\r\n\r\n    // Some pdf's text items are not in order. This is most likely a result of creating it\r\n    // from design softwares, e.g. canvas. The commented out method can sort pageTextItems\r\n    // by y position to put them back in order. But it is not used since it might be more\r\n    // helpful to let users know that the pdf is not in order.\r\n    // pageTextItems.sort((a, b) => Math.round(b.y) - Math.round(a.y));\r\n\r\n    // Add text items of each page to total\r\n    textItems.push(...pageTextItems);\r\n  }\r\n\r\n  // Filter out empty space textItem noise\r\n  const isEmptySpace = (textItem: TextItem) =>\r\n    !textItem.hasEOL && textItem.text.trim() === \"\";\r\n  textItems = textItems.filter((textItem) => !isEmptySpace(textItem));\r\n\r\n  return textItems;\r\n};\r\n","import type { Lines, TextItem } from \"../../../../lib/parse-resume-from-pdf/types\";\r\n\r\n/**\r\n * List of bullet points\r\n * Reference: https://stackoverflow.com/questions/56540160/why-isnt-there-a-medium-small-black-circle-in-unicode\r\n * U+22C5   DOT OPERATOR (⋅)\r\n * U+2219   BULLET OPERATOR (∙)\r\n * U+1F784  BLACK SLIGHTLY SMALL CIRCLE (🞄)\r\n * U+2022   BULLET (•) -------- most common\r\n * U+2981   Z NOTATION SPOT (⦁)\r\n * U+26AB   MEDIUM BLACK CIRCLE (⚫︎)\r\n * U+25CF   BLACK CIRCLE (●)\r\n * U+2B24   BLACK LARGE CIRCLE (⬤)\r\n * U+26AC   MEDIUM SMALL WHITE CIRCLE ⚬\r\n * U+25CB   WHITE CIRCLE ○\r\n */\r\nexport const BULLET_POINTS = [\r\n  \"⋅\",\r\n  \"∙\",\r\n  \"🞄\",\r\n  \"•\",\r\n  \"⦁\",\r\n  \"⚫︎\",\r\n  \"●\",\r\n  \"⬤\",\r\n  \"⚬\",\r\n  \"○\",\r\n];\r\n\r\n/**\r\n * Convert bullet point lines into a string array aka descriptions.\r\n */\r\nexport const getBulletPointsFromLines = (lines: Lines): string[] => {\r\n  // Simply return all lines with text item joined together if there is no bullet point\r\n  const firstBulletPointLineIndex = getFirstBulletPointLineIdx(lines);\r\n  if (firstBulletPointLineIndex === undefined) {\r\n    return lines.map((line) => line.map((item) => item.text).join(\" \"));\r\n  }\r\n\r\n  // Otherwise, process and remove bullet points\r\n\r\n  // Combine all lines into a single string\r\n  let lineStr = \"\";\r\n  for (let item of lines.flat()) {\r\n    const text = item.text;\r\n    // Make sure a space is added between 2 words\r\n    if (!lineStr.endsWith(\" \") && !text.startsWith(\" \")) {\r\n      lineStr += \" \";\r\n    }\r\n    lineStr += text;\r\n  }\r\n\r\n  // Get the most common bullet point\r\n  const commonBulletPoint = getMostCommonBulletPoint(lineStr);\r\n\r\n  // Start line string from the beginning of the first bullet point\r\n  const firstBulletPointIndex = lineStr.indexOf(commonBulletPoint);\r\n  if (firstBulletPointIndex !== -1) {\r\n    lineStr = lineStr.slice(firstBulletPointIndex);\r\n  }\r\n\r\n  // Divide the single string using bullet point as divider\r\n  return lineStr\r\n    .split(commonBulletPoint)\r\n    .map((text) => text.trim())\r\n    .filter((text) => !!text);\r\n};\r\n\r\nconst getMostCommonBulletPoint = (str: string): string => {\r\n  const bulletToCount: { [bullet: string]: number } = BULLET_POINTS.reduce(\r\n    (acc: { [bullet: string]: number }, cur) => {\r\n      acc[cur] = 0;\r\n      return acc;\r\n    },\r\n    {}\r\n  );\r\n  let bulletWithMostCount = BULLET_POINTS[0];\r\n  let bulletMaxCount = 0;\r\n  for (let char of str) {\r\n    if (bulletToCount.hasOwnProperty(char)) {\r\n      bulletToCount[char]++;\r\n      if (bulletToCount[char] > bulletMaxCount) {\r\n        bulletWithMostCount = char;\r\n      }\r\n    }\r\n  }\r\n  return bulletWithMostCount;\r\n};\r\n\r\nconst getFirstBulletPointLineIdx = (lines: Lines): number | undefined => {\r\n  for (let i = 0; i < lines.length; i++) {\r\n    for (let item of lines[i]) {\r\n      if (BULLET_POINTS.some((bullet) => item.text.includes(bullet))) {\r\n        return i;\r\n      }\r\n    }\r\n  }\r\n  return undefined;\r\n};\r\n\r\n// Only consider words that don't contain numbers\r\nconst isWord = (str: string) => /^[^0-9]+$/.test(str);\r\nconst hasAtLeast8Words = (item: TextItem) =>\r\n  item.text.split(/\\s/).filter(isWord).length >= 8;\r\n\r\nexport const getDescriptionsLineIdx = (lines: Lines): number | undefined => {\r\n  // The main heuristic to determine descriptions is to check if has bullet point\r\n  let idx = getFirstBulletPointLineIdx(lines);\r\n\r\n  // Fallback heuristic if the main heuristic doesn't apply (e.g. LinkedIn resume) to\r\n  // check if the line has at least 8 words\r\n  if (idx === undefined) {\r\n    for (let i = 0; i < lines.length; i++) {\r\n      const line = lines[i];\r\n      if (line.length === 1 && hasAtLeast8Words(line[0])) {\r\n        idx = i;\r\n        break;\r\n      }\r\n    }\r\n  }\r\n\r\n  return idx;\r\n};\r\n","import { BULLET_POINTS } from \"../../lib/parse-resume-from-pdf/extract-resume-from-sections/lib/bullet-points\";\r\nimport type { TextItems, Line, Lines } from \"../../lib/parse-resume-from-pdf/types\";\r\n\r\n/**\r\n * Step 2: Group text items into lines. This returns an array where each position\r\n * contains text items in the same line of the pdf file.\r\n */\r\nexport const groupTextItemsIntoLines = (textItems: TextItems): Lines => {\r\n  const lines: Lines = [];\r\n\r\n  // Group text items into lines based on hasEOL\r\n  let line: Line = [];\r\n  for (let item of textItems) {\r\n    // If item is EOL, add current line to lines and start a new empty line\r\n    if (item.hasEOL) {\r\n      if (item.text.trim() !== \"\") {\r\n        line.push({ ...item });\r\n      }\r\n      lines.push(line);\r\n      line = [];\r\n    }\r\n    // Otherwise, add item to current line\r\n    else if (item.text.trim() !== \"\") {\r\n      line.push({ ...item });\r\n    }\r\n  }\r\n  // Add last line if there is item in last line\r\n  if (line.length > 0) {\r\n    lines.push(line);\r\n  }\r\n\r\n  // Many pdf docs are not well formatted, e.g. due to converting from other docs.\r\n  // This creates many noises, where a single text item is divided into multiple\r\n  // ones. This step is to merge adjacent text items if their distance is smaller\r\n  // than a typical char width to filter out those noises.\r\n  const typicalCharWidth = getTypicalCharWidth(lines.flat());\r\n  for (let line of lines) {\r\n    // Start from the end of the line to make things easier to merge and delete\r\n    for (let i = line.length - 1; i > 0; i--) {\r\n      const currentItem = line[i];\r\n      const leftItem = line[i - 1];\r\n      const leftItemXEnd = leftItem.x + leftItem.width;\r\n      const distance = currentItem.x - leftItemXEnd;\r\n      if (distance <= typicalCharWidth) {\r\n        if (shouldAddSpaceBetweenText(leftItem.text, currentItem.text)) {\r\n          leftItem.text += \" \";\r\n        }\r\n        leftItem.text += currentItem.text;\r\n        // Update leftItem width to include currentItem after merge before deleting current item\r\n        const currentItemXEnd = currentItem.x + currentItem.width;\r\n        leftItem.width = currentItemXEnd - leftItem.x;\r\n        line.splice(i, 1);\r\n      }\r\n    }\r\n  }\r\n\r\n  return lines;\r\n};\r\n\r\n// Sometimes a space is lost while merging adjacent text items. This accounts for some of those cases\r\nconst shouldAddSpaceBetweenText = (leftText: string, rightText: string) => {\r\n  const leftTextEnd = leftText[leftText.length - 1];\r\n  const rightTextStart = rightText[0];\r\n  const conditions = [\r\n    [\":\", \",\", \"|\", \".\", ...BULLET_POINTS].includes(leftTextEnd) &&\r\n      rightTextStart !== \" \",\r\n    leftTextEnd !== \" \" && [\"|\", ...BULLET_POINTS].includes(rightTextStart),\r\n  ];\r\n\r\n  return conditions.some((condition) => condition);\r\n};\r\n\r\n/**\r\n * Return the width of a typical character. (Helper util for groupTextItemsIntoLines)\r\n *\r\n * A pdf file uses different characters, each with different width due to different\r\n * font family and font size. This util first extracts the most typically used font\r\n * family and font height, and compute the average character width using text items\r\n * that match the typical font family and height.\r\n */\r\nconst getTypicalCharWidth = (textItems: TextItems): number => {\r\n  // Exclude empty space \" \" in calculations since its width isn't precise\r\n  textItems = textItems.filter((item) => item.text.trim() !== \"\");\r\n\r\n  const heightToCount: { [height: number]: number } = {};\r\n  let commonHeight = 0;\r\n  let heightMaxCount = 0;\r\n\r\n  const fontNameToCount: { [fontName: string]: number } = {};\r\n  let commonFontName = \"\";\r\n  let fontNameMaxCount = 0;\r\n\r\n  for (let item of textItems) {\r\n    const { text, height, fontName } = item;\r\n    // Process height\r\n    if (!heightToCount[height]) {\r\n      heightToCount[height] = 0;\r\n    }\r\n    heightToCount[height]++;\r\n    if (heightToCount[height] > heightMaxCount) {\r\n      commonHeight = height;\r\n      heightMaxCount = heightToCount[height];\r\n    }\r\n\r\n    // Process font name\r\n    if (!fontNameToCount[fontName]) {\r\n      fontNameToCount[fontName] = 0;\r\n    }\r\n    fontNameToCount[fontName] += text.length;\r\n    if (fontNameToCount[fontName] > fontNameMaxCount) {\r\n      commonFontName = fontName;\r\n      fontNameMaxCount = fontNameToCount[fontName];\r\n    }\r\n  }\r\n\r\n  // Find the text items that match common font family and height\r\n  const commonTextItems = textItems.filter(\r\n    (item) => item.fontName === commonFontName && item.height === commonHeight\r\n  );\r\n  // Aggregate total width and number of characters of all common text items\r\n  const [totalWidth, numChars] = commonTextItems.reduce(\r\n    (acc, cur) => {\r\n      const [preWidth, prevChars] = acc;\r\n      return [preWidth + cur.width, prevChars + cur.text.length];\r\n    },\r\n    [0, 0]\r\n  );\r\n  const typicalCharWidth = totalWidth / numChars;\r\n\r\n  return typicalCharWidth;\r\n};\r\n","import type { TextItem, FeatureSet } from \"../../../../lib/parse-resume-from-pdf/types\";\r\n\r\nconst isTextItemBold = (fontName: string) =>\r\n  fontName.toLowerCase().includes(\"bold\");\r\nexport const isBold = (item: TextItem) => isTextItemBold(item.fontName);\r\nexport const hasLetter = (item: TextItem) => /[a-zA-Z]/.test(item.text);\r\nexport const hasNumber = (item: TextItem) => /[0-9]/.test(item.text);\r\nexport const hasComma = (item: TextItem) => item.text.includes(\",\");\r\nexport const getHasText = (text: string) => (item: TextItem) =>\r\n  item.text.includes(text);\r\nexport const hasOnlyLettersSpacesAmpersands = (item: TextItem) =>\r\n  /^[A-Za-z\\s&]+$/.test(item.text);\r\nexport const hasLetterAndIsAllUpperCase = (item: TextItem) =>\r\n  hasLetter(item) && item.text.toUpperCase() === item.text;\r\n\r\n// Date Features\r\nconst hasYear = (item: TextItem) => /(?:19|20)\\d{2}/.test(item.text);\r\n// prettier-ignore\r\nconst MONTHS = ['January', 'February', 'March', 'April', 'May', 'June', 'July', 'August', 'September', 'October', 'November', 'December'];\r\nconst hasMonth = (item: TextItem) =>\r\n  MONTHS.some(\r\n    (month) =>\r\n      item.text.includes(month) || item.text.includes(month.slice(0, 4))\r\n  );\r\nconst SEASONS = [\"Summer\", \"Fall\", \"Spring\", \"Winter\"];\r\nconst hasSeason = (item: TextItem) =>\r\n  SEASONS.some((season) => item.text.includes(season));\r\nconst hasPresent = (item: TextItem) => item.text.includes(\"Present\");\r\nexport const DATE_FEATURE_SETS: FeatureSet[] = [\r\n  [hasYear, 1],\r\n  [hasMonth, 1],\r\n  [hasSeason, 1],\r\n  [hasPresent, 1],\r\n  [hasComma, -1],\r\n];\r\n","import type { ResumeKey } from \"../../lib/redux/types\";\r\nimport type {\r\n  Line,\r\n  Lines,\r\n  ResumeSectionToLines,\r\n} from \"../../lib/parse-resume-from-pdf/types\";\r\nimport {\r\n  hasLetterAndIsAllUpperCase,\r\n  hasOnlyLettersSpacesAmpersands,\r\n  isBold,\r\n} from \"../../lib/parse-resume-from-pdf/extract-resume-from-sections/lib/common-features\";\r\n\r\nexport const PROFILE_SECTION: ResumeKey = \"profile\";\r\n\r\n/**\r\n * Step 3. Group lines into sections\r\n *\r\n * Every section (except the profile section) starts with a section title that\r\n * takes up the entire line. This is a common pattern not just in resumes but\r\n * also in books and blogs. The resume parser uses this pattern to group lines\r\n * into the closest section title above these lines.\r\n */\r\nexport const groupLinesIntoSections = (lines: Lines) => {\r\n  let sections: ResumeSectionToLines = {};\r\n  let sectionName: string = PROFILE_SECTION;\r\n  let sectionLines = [];\r\n  for (let i = 0; i < lines.length; i++) {\r\n    const line = lines[i];\r\n    const text = line[0]?.text.trim();\r\n    if (isSectionTitle(line, i)) {\r\n      sections[sectionName] = [...sectionLines];\r\n      sectionName = text;\r\n      sectionLines = [];\r\n    } else {\r\n      sectionLines.push(line);\r\n    }\r\n  }\r\n  if (sectionLines.length > 0) {\r\n    sections[sectionName] = [...sectionLines];\r\n  }\r\n  return sections;\r\n};\r\n\r\nconst SECTION_TITLE_PRIMARY_KEYWORDS = [\r\n  \"experience\",\r\n  \"education\",\r\n  \"project\",\r\n  \"skill\",\r\n];\r\nconst SECTION_TITLE_SECONDARY_KEYWORDS = [\r\n  \"job\",\r\n  \"course\",\r\n  \"extracurricular\",\r\n  \"objective\",\r\n  \"summary\", // LinkedIn generated resume has a summary section\r\n  \"award\",\r\n  \"honor\",\r\n  \"project\",\r\n];\r\nconst SECTION_TITLE_KEYWORDS = [\r\n  ...SECTION_TITLE_PRIMARY_KEYWORDS,\r\n  ...SECTION_TITLE_SECONDARY_KEYWORDS,\r\n];\r\n\r\nconst isSectionTitle = (line: Line, lineNumber: number) => {\r\n  const isFirstTwoLines = lineNumber < 2;\r\n  const hasMoreThanOneItemInLine = line.length > 1;\r\n  const hasNoItemInLine = line.length === 0;\r\n  if (isFirstTwoLines || hasMoreThanOneItemInLine || hasNoItemInLine) {\r\n    return false;\r\n  }\r\n\r\n  const textItem = line[0];\r\n\r\n  // The main heuristic to determine a section title is to check if the text is double emphasized\r\n  // to be both bold and all uppercase, which is generally true for a well formatted resume\r\n  if (isBold(textItem) && hasLetterAndIsAllUpperCase(textItem)) {\r\n    return true;\r\n  }\r\n\r\n  // The following is a fallback heuristic to detect section title if it includes a keyword match\r\n  // (This heuristics is not well tested and may not work well)\r\n  const text = textItem.text.trim();\r\n  const textHasAtMost2Words =\r\n    text.split(\" \").filter((s) => s !== \"&\").length <= 2;\r\n  const startsWithCapitalLetter = /[A-Z]/.test(text.slice(0, 1));\r\n\r\n  if (\r\n    textHasAtMost2Words &&\r\n    hasOnlyLettersSpacesAmpersands(textItem) &&\r\n    startsWithCapitalLetter &&\r\n    SECTION_TITLE_KEYWORDS.some((keyword) =>\r\n      text.toLowerCase().includes(keyword)\r\n    )\r\n  ) {\r\n    return true;\r\n  }\r\n\r\n  return false;\r\n};\r\n","import type { ResumeSectionToLines } from \"../../../../lib/parse-resume-from-pdf/types\";\r\n\r\n/**\r\n * Return section lines that contain any of the keywords.\r\n */\r\nexport const getSectionLinesByKeywords = (\r\n  sections: ResumeSectionToLines,\r\n  keywords: string[]\r\n) => {\r\n  for (const sectionName in sections) {\r\n    const hasKeyWord = keywords.some((keyword) =>\r\n      sectionName.toLowerCase().includes(keyword)\r\n    );\r\n    if (hasKeyWord) {\r\n      return sections[sectionName];\r\n    }\r\n  }\r\n  return [];\r\n};\r\n","import type {\r\n  TextItems,\r\n  TextScores,\r\n  FeatureSet,\r\n} from \"../../../../lib/parse-resume-from-pdf/types\";\r\n\r\nconst computeFeatureScores = (\r\n  textItems: TextItems,\r\n  featureSets: FeatureSet[]\r\n): TextScores => {\r\n  const textScores = textItems.map((item) => ({\r\n    text: item.text,\r\n    score: 0,\r\n    match: false,\r\n  }));\r\n\r\n  for (let i = 0; i < textItems.length; i++) {\r\n    const textItem = textItems[i];\r\n\r\n    for (const featureSet of featureSets) {\r\n      const [hasFeature, score, returnMatchingText] = featureSet;\r\n      const result = hasFeature(textItem);\r\n      if (result) {\r\n        let text = textItem.text;\r\n        if (returnMatchingText && typeof result === \"object\") {\r\n          text = result[0];\r\n        }\r\n\r\n        const textScore = textScores[i];\r\n        if (textItem.text === text) {\r\n          textScore.score += score;\r\n          if (returnMatchingText) {\r\n            textScore.match = true;\r\n          }\r\n        } else {\r\n          textScores.push({ text, score, match: true });\r\n        }\r\n      }\r\n    }\r\n  }\r\n  return textScores;\r\n};\r\n\r\n/**\r\n * Core util for the feature scoring system.\r\n *\r\n * It runs each text item through all feature sets and sums up the matching feature scores.\r\n * It then returns the text item with the highest computed feature score.\r\n */\r\nexport const getTextWithHighestFeatureScore = (\r\n  textItems: TextItems,\r\n  featureSets: FeatureSet[],\r\n  returnEmptyStringIfHighestScoreIsNotPositive = true,\r\n  returnConcatenatedStringForTextsWithSameHighestScore = false\r\n) => {\r\n  const textScores = computeFeatureScores(textItems, featureSets);\r\n\r\n  let textsWithHighestFeatureScore: string[] = [];\r\n  let highestScore = -Infinity;\r\n  for (const { text, score } of textScores) {\r\n    if (score >= highestScore) {\r\n      if (score > highestScore) {\r\n        textsWithHighestFeatureScore = [];\r\n      }\r\n      textsWithHighestFeatureScore.push(text);\r\n      highestScore = score;\r\n    }\r\n  }\r\n\r\n  if (returnEmptyStringIfHighestScoreIsNotPositive && highestScore <= 0)\r\n    return [\"\", textScores] as const;\r\n\r\n  // Note: If textItems is an empty array, textsWithHighestFeatureScore[0] is undefined, so we default it to empty string\r\n  const text = !returnConcatenatedStringForTextsWithSameHighestScore\r\n    ? textsWithHighestFeatureScore[0] ?? \"\"\r\n    : textsWithHighestFeatureScore.map((s) => s.trim()).join(\" \");\r\n\r\n  return [text, textScores] as const;\r\n};\r\n","import type {\r\n  ResumeSectionToLines,\r\n  TextItem,\r\n  FeatureSet,\r\n} from \"../../../lib/parse-resume-from-pdf/types\";\r\nimport { getSectionLinesByKeywords } from \"../../../lib/parse-resume-from-pdf/extract-resume-from-sections/lib/get-section-lines\";\r\nimport {\r\n  isBold,\r\n  hasNumber,\r\n  hasComma,\r\n  hasLetter,\r\n  hasLetterAndIsAllUpperCase,\r\n} from \"../../../lib/parse-resume-from-pdf/extract-resume-from-sections/lib/common-features\";\r\nimport { getTextWithHighestFeatureScore } from \"../../../lib/parse-resume-from-pdf/extract-resume-from-sections/lib/feature-scoring-system\";\r\n\r\n// Name\r\nexport const matchOnlyLetterSpaceOrPeriod = (item: TextItem) =>\r\n  item.text.match(/^[a-zA-Z\\s\\.]+$/);\r\n\r\n// Email\r\n// Simple email regex: xxx@xxx.xxx (xxx = anything not space)\r\nexport const matchEmail = (item: TextItem) => item.text.match(/\\S+@\\S+\\.\\S+/);\r\nconst hasAt = (item: TextItem) => item.text.includes(\"@\");\r\n\r\n// Phone\r\n// Simple phone regex that matches (xxx)-xxx-xxxx where () and - are optional, - can also be space\r\nexport const matchPhone = (item: TextItem) =>\r\n  item.text.match(/\\(?\\d{3}\\)?[\\s-]?\\d{3}[\\s-]?\\d{4}/);\r\nconst hasParenthesis = (item: TextItem) => /\\([0-9]+\\)/.test(item.text);\r\n\r\n// Location\r\n// Simple location regex that matches \"<City>, <ST>\"\r\nexport const matchCityAndState = (item: TextItem) =>\r\n  item.text.match(/[A-Z][a-zA-Z\\s]+, [A-Z]{2}/);\r\n\r\n// Url\r\n// Simple url regex that matches \"xxx.xxx/xxx\" (xxx = anything not space)\r\nexport const matchUrl = (item: TextItem) => item.text.match(/\\S+\\.[a-z]+\\/\\S+/);\r\n// Match https://xxx.xxx where s is optional\r\nconst matchUrlHttpFallback = (item: TextItem) =>\r\n  item.text.match(/https?:\\/\\/\\S+\\.\\S+/);\r\n// Match www.xxx.xxx\r\nconst matchUrlWwwFallback = (item: TextItem) =>\r\n  item.text.match(/www\\.\\S+\\.\\S+/);\r\nconst hasSlash = (item: TextItem) => item.text.includes(\"/\");\r\n\r\n// Summary\r\nconst has4OrMoreWords = (item: TextItem) => item.text.split(\" \").length >= 4;\r\n\r\n/**\r\n *              Unique Attribute\r\n * Name         Bold or Has all uppercase letter\r\n * Email        Has @\r\n * Phone        Has ()\r\n * Location     Has ,    (overlap with summary)\r\n * Url          Has slash\r\n * Summary      Has 4 or more words\r\n */\r\n\r\n/**\r\n * Name -> contains only letters/space/period, e.g. Leonardo W. DiCaprio\r\n *         (it isn't common to include middle initial in resume)\r\n *      -> is bolded or has all letters as uppercase\r\n */\r\nconst NAME_FEATURE_SETS: FeatureSet[] = [\r\n  [matchOnlyLetterSpaceOrPeriod, 3, true],\r\n  [isBold, 2],\r\n  [hasLetterAndIsAllUpperCase, 2],\r\n  // Match against other unique attributes\r\n  [hasAt, -4], // Email\r\n  [hasNumber, -4], // Phone\r\n  [hasParenthesis, -4], // Phone\r\n  [hasComma, -4], // Location\r\n  [hasSlash, -4], // Url\r\n  [has4OrMoreWords, -2], // Summary\r\n];\r\n\r\n// Email -> match email regex xxx@xxx.xxx\r\nconst EMAIL_FEATURE_SETS: FeatureSet[] = [\r\n  [matchEmail, 4, true],\r\n  [isBold, -1], // Name\r\n  [hasLetterAndIsAllUpperCase, -1], // Name\r\n  [hasParenthesis, -4], // Phone\r\n  [hasComma, -4], // Location\r\n  [hasSlash, -4], // Url\r\n  [has4OrMoreWords, -4], // Summary\r\n];\r\n\r\n// Phone -> match phone regex (xxx)-xxx-xxxx\r\nconst PHONE_FEATURE_SETS: FeatureSet[] = [\r\n  [matchPhone, 4, true],\r\n  [hasLetter, -4], // Name, Email, Location, Url, Summary\r\n];\r\n\r\n// Location -> match location regex <City>, <ST>\r\nconst LOCATION_FEATURE_SETS: FeatureSet[] = [\r\n  [matchCityAndState, 4, true],\r\n  [isBold, -1], // Name\r\n  [hasAt, -4], // Email\r\n  [hasParenthesis, -3], // Phone\r\n  [hasSlash, -4], // Url\r\n];\r\n\r\n// URL -> match url regex xxx.xxx/xxx\r\nconst URL_FEATURE_SETS: FeatureSet[] = [\r\n  [matchUrl, 4, true],\r\n  [matchUrlHttpFallback, 3, true],\r\n  [matchUrlWwwFallback, 3, true],\r\n  [isBold, -1], // Name\r\n  [hasAt, -4], // Email\r\n  [hasParenthesis, -3], // Phone\r\n  [hasComma, -4], // Location\r\n  [has4OrMoreWords, -4], // Summary\r\n];\r\n\r\n// Summary -> has 4 or more words\r\nconst SUMMARY_FEATURE_SETS: FeatureSet[] = [\r\n  [has4OrMoreWords, 4],\r\n  [isBold, -1], // Name\r\n  [hasAt, -4], // Email\r\n  [hasParenthesis, -3], // Phone\r\n  [matchCityAndState, -4, false], // Location\r\n];\r\n\r\nexport const extractProfile = (sections: ResumeSectionToLines) => {\r\n  const lines = sections.profile || [];\r\n  const textItems = lines.flat();\r\n\r\n  const [name, nameScores] = getTextWithHighestFeatureScore(\r\n    textItems,\r\n    NAME_FEATURE_SETS\r\n  );\r\n  const [email, emailScores] = getTextWithHighestFeatureScore(\r\n    textItems,\r\n    EMAIL_FEATURE_SETS\r\n  );\r\n  const [phone, phoneScores] = getTextWithHighestFeatureScore(\r\n    textItems,\r\n    PHONE_FEATURE_SETS\r\n  );\r\n  const [location, locationScores] = getTextWithHighestFeatureScore(\r\n    textItems,\r\n    LOCATION_FEATURE_SETS\r\n  );\r\n  const [url, urlScores] = getTextWithHighestFeatureScore(\r\n    textItems,\r\n    URL_FEATURE_SETS\r\n  );\r\n  const [summary, summaryScores] = getTextWithHighestFeatureScore(\r\n    textItems,\r\n    SUMMARY_FEATURE_SETS,\r\n    undefined,\r\n    true\r\n  );\r\n\r\n  const summaryLines = getSectionLinesByKeywords(sections, [\"summary\"]);\r\n  const summarySection = summaryLines\r\n    .flat()\r\n    .map((textItem) => textItem.text)\r\n    .join(\" \");\r\n  const objectiveLines = getSectionLinesByKeywords(sections, [\"objective\"]);\r\n  const objectiveSection = objectiveLines\r\n    .flat()\r\n    .map((textItem) => textItem.text)\r\n    .join(\" \");\r\n\r\n  return {\r\n    profile: {\r\n      name,\r\n      email,\r\n      phone,\r\n      location,\r\n      url,\r\n      // Dedicated section takes higher precedence over profile summary\r\n      summary: summarySection || objectiveSection || summary,\r\n    },\r\n    // For debugging\r\n    profileScores: {\r\n      name: nameScores,\r\n      email: emailScores,\r\n      phone: phoneScores,\r\n      location: locationScores,\r\n      url: urlScores,\r\n      summary: summaryScores,\r\n    },\r\n  };\r\n};\r\n","import { BULLET_POINTS } from \"../../../../lib/parse-resume-from-pdf/extract-resume-from-sections/lib/bullet-points\";\r\nimport { isBold } from \"../../../../lib/parse-resume-from-pdf/extract-resume-from-sections/lib/common-features\";\r\nimport type { Lines, Line, Subsections } from \"../../../../lib/parse-resume-from-pdf/types\";\r\n\r\n/**\r\n * Divide lines into subsections based on difference in line gap or bold text.\r\n *\r\n * For profile section, we can directly pass all the text items to the feature\r\n * scoring systems. But for other sections, such as education and work experience,\r\n * we have to first divide the section into subsections since there can be multiple\r\n * schools or work experiences in the section. The feature scoring system then\r\n * process each subsection to retrieve each's resume attributes and append the results.\r\n */\r\nexport const divideSectionIntoSubsections = (lines: Lines): Subsections => {\r\n  // The main heuristic to determine a subsection is to check if its vertical line gap\r\n  // is larger than the typical line gap * 1.4\r\n  const isLineNewSubsectionByLineGap =\r\n    createIsLineNewSubsectionByLineGap(lines);\r\n\r\n  let subsections = createSubsections(lines, isLineNewSubsectionByLineGap);\r\n\r\n  // Fallback heuristic if the main heuristic doesn't apply to check if the text item is bolded\r\n  if (subsections.length === 1) {\r\n    const isLineNewSubsectionByBold = (line: Line, prevLine: Line) => {\r\n      if (\r\n        !isBold(prevLine[0]) &&\r\n        isBold(line[0]) &&\r\n        // Ignore bullet points that sometimes being marked as bolded\r\n        !BULLET_POINTS.includes(line[0].text)\r\n      ) {\r\n        return true;\r\n      }\r\n      return false;\r\n    };\r\n\r\n    subsections = createSubsections(lines, isLineNewSubsectionByBold);\r\n  }\r\n\r\n  return subsections;\r\n};\r\n\r\ntype IsLineNewSubsection = (line: Line, prevLine: Line) => boolean;\r\n\r\nconst createIsLineNewSubsectionByLineGap = (\r\n  lines: Lines\r\n): IsLineNewSubsection => {\r\n  // Extract the common typical line gap\r\n  const lineGapToCount: { [lineGap: number]: number } = {};\r\n  const linesY = lines.map((line) => line[0].y);\r\n  let lineGapWithMostCount: number = 0;\r\n  let maxCount = 0;\r\n  for (let i = 1; i < linesY.length; i++) {\r\n    const lineGap = Math.round(linesY[i - 1] - linesY[i]);\r\n    if (!lineGapToCount[lineGap]) lineGapToCount[lineGap] = 0;\r\n    lineGapToCount[lineGap] += 1;\r\n    if (lineGapToCount[lineGap] > maxCount) {\r\n      lineGapWithMostCount = lineGap;\r\n      maxCount = lineGapToCount[lineGap];\r\n    }\r\n  }\r\n  // Use common line gap to set a sub section threshold\r\n  const subsectionLineGapThreshold = lineGapWithMostCount * 1.4;\r\n\r\n  const isLineNewSubsection = (line: Line, prevLine: Line) => {\r\n    return Math.round(prevLine[0].y - line[0].y) > subsectionLineGapThreshold;\r\n  };\r\n\r\n  return isLineNewSubsection;\r\n};\r\n\r\nconst createSubsections = (\r\n  lines: Lines,\r\n  isLineNewSubsection: IsLineNewSubsection\r\n): Subsections => {\r\n  const subsections: Subsections = [];\r\n  let subsection: Lines = [];\r\n  for (let i = 0; i < lines.length; i++) {\r\n    const line = lines[i];\r\n    if (i === 0) {\r\n      subsection.push(line);\r\n      continue;\r\n    }\r\n    if (isLineNewSubsection(line, lines[i - 1])) {\r\n      subsections.push(subsection);\r\n      subsection = [];\r\n    }\r\n    subsection.push(line);\r\n  }\r\n  if (subsection.length > 0) {\r\n    subsections.push(subsection);\r\n  }\r\n  return subsections;\r\n};\r\n","import type {\r\n  TextItem,\r\n  FeatureSet,\r\n  ResumeSectionToLines,\r\n} from \"../../../lib/parse-resume-from-pdf/types\";\r\nimport type { ResumeEducation } from \"../../../lib/redux/types\";\r\nimport { getSectionLinesByKeywords } from \"../../../lib/parse-resume-from-pdf/extract-resume-from-sections/lib/get-section-lines\";\r\nimport { divideSectionIntoSubsections } from \"../../../lib/parse-resume-from-pdf/extract-resume-from-sections/lib/subsections\";\r\nimport {\r\n  DATE_FEATURE_SETS,\r\n  hasComma,\r\n  hasLetter,\r\n  hasNumber,\r\n} from \"../../../lib/parse-resume-from-pdf/extract-resume-from-sections/lib/common-features\";\r\nimport { getTextWithHighestFeatureScore } from \"../../../lib/parse-resume-from-pdf/extract-resume-from-sections/lib/feature-scoring-system\";\r\nimport {\r\n  getBulletPointsFromLines,\r\n  getDescriptionsLineIdx,\r\n} from \"../../../lib/parse-resume-from-pdf/extract-resume-from-sections/lib/bullet-points\";\r\n\r\n/**\r\n *              Unique Attribute\r\n * School       Has school\r\n * Degree       Has degree\r\n * GPA          Has number\r\n */\r\n\r\n// prettier-ignore\r\nconst SCHOOLS = ['College', 'University', 'Institute', 'School', 'Academy', 'BASIS', 'Magnet']\r\nconst hasSchool = (item: TextItem) =>\r\n  SCHOOLS.some((school) => item.text.includes(school));\r\n// prettier-ignore\r\nconst DEGREES = [\"Associate\", \"Bachelor\", \"Master\", \"PhD\", \"Ph.\"];\r\nconst hasDegree = (item: TextItem) =>\r\n  DEGREES.some((degree) => item.text.includes(degree)) ||\r\n  /[ABM][A-Z\\.]/.test(item.text); // Match AA, B.S., MBA, etc.\r\nconst matchGPA = (item: TextItem) => item.text.match(/[0-4]\\.\\d{1,2}/);\r\nconst matchGrade = (item: TextItem) => {\r\n  const grade = parseFloat(item.text);\r\n  if (Number.isFinite(grade) && grade <= 110) {\r\n    return [String(grade)] as RegExpMatchArray;\r\n  }\r\n  return null;\r\n};\r\n\r\nconst SCHOOL_FEATURE_SETS: FeatureSet[] = [\r\n  [hasSchool, 4],\r\n  [hasDegree, -4],\r\n  [hasNumber, -4],\r\n];\r\n\r\nconst DEGREE_FEATURE_SETS: FeatureSet[] = [\r\n  [hasDegree, 4],\r\n  [hasSchool, -4],\r\n  [hasNumber, -3],\r\n];\r\n\r\nconst GPA_FEATURE_SETS: FeatureSet[] = [\r\n  [matchGPA, 4, true],\r\n  [matchGrade, 3, true],\r\n  [hasComma, -3],\r\n  [hasLetter, -4],\r\n];\r\n\r\nexport const extractEducation = (sections: ResumeSectionToLines) => {\r\n  const educations: ResumeEducation[] = [];\r\n  const educationsScores = [];\r\n  const lines = getSectionLinesByKeywords(sections, [\"education\"]);\r\n  const subsections = divideSectionIntoSubsections(lines);\r\n  for (const subsectionLines of subsections) {\r\n    const textItems = subsectionLines.flat();\r\n    const [school, schoolScores] = getTextWithHighestFeatureScore(\r\n      textItems,\r\n      SCHOOL_FEATURE_SETS\r\n    );\r\n    const [degree, degreeScores] = getTextWithHighestFeatureScore(\r\n      textItems,\r\n      DEGREE_FEATURE_SETS\r\n    );\r\n    const [gpa, gpaScores] = getTextWithHighestFeatureScore(\r\n      textItems,\r\n      GPA_FEATURE_SETS\r\n    );\r\n    const [date, dateScores] = getTextWithHighestFeatureScore(\r\n      textItems,\r\n      DATE_FEATURE_SETS\r\n    );\r\n\r\n    let descriptions: string[] = [];\r\n    const descriptionsLineIdx = getDescriptionsLineIdx(subsectionLines);\r\n    if (descriptionsLineIdx !== undefined) {\r\n      const descriptionsLines = subsectionLines.slice(descriptionsLineIdx);\r\n      descriptions = getBulletPointsFromLines(descriptionsLines);\r\n    }\r\n\r\n    educations.push({ school, degree, gpa, date, descriptions });\r\n    educationsScores.push({\r\n      schoolScores,\r\n      degreeScores,\r\n      gpaScores,\r\n      dateScores,\r\n    });\r\n  }\r\n\r\n  if (educations.length !== 0) {\r\n    const coursesLines = getSectionLinesByKeywords(sections, [\"course\"]);\r\n    if (coursesLines.length !== 0) {\r\n      educations[0].descriptions.push(\r\n        \"Courses: \" +\r\n          coursesLines\r\n            .flat()\r\n            .map((item) => item.text)\r\n            .join(\" \")\r\n      );\r\n    }\r\n  }\r\n\r\n  return {\r\n    educations,\r\n    educationsScores,\r\n  };\r\n};\r\n","import type { ResumeWorkExperience } from \"../../../lib/redux/types\";\r\nimport type {\r\n  TextItem,\r\n  FeatureSet,\r\n  ResumeSectionToLines,\r\n} from \"../../../lib/parse-resume-from-pdf/types\";\r\nimport { getSectionLinesByKeywords } from \"../../../lib/parse-resume-from-pdf/extract-resume-from-sections/lib/get-section-lines\";\r\nimport {\r\n  DATE_FEATURE_SETS,\r\n  hasNumber,\r\n  getHasText,\r\n  isBold,\r\n} from \"../../../lib/parse-resume-from-pdf/extract-resume-from-sections/lib/common-features\";\r\nimport { divideSectionIntoSubsections } from \"../../../lib/parse-resume-from-pdf/extract-resume-from-sections/lib/subsections\";\r\nimport { getTextWithHighestFeatureScore } from \"../../../lib/parse-resume-from-pdf/extract-resume-from-sections/lib/feature-scoring-system\";\r\nimport {\r\n  getBulletPointsFromLines,\r\n  getDescriptionsLineIdx,\r\n} from \"../../../lib/parse-resume-from-pdf/extract-resume-from-sections/lib/bullet-points\";\r\n\r\n// prettier-ignore\r\nconst WORK_EXPERIENCE_KEYWORDS_LOWERCASE = ['work', 'experience', 'employment', 'history', 'job'];\r\n// prettier-ignore\r\nconst JOB_TITLES = ['Accountant', 'Administrator', 'Advisor', 'Agent', 'Analyst', 'Apprentice', 'Architect', 'Assistant', 'Associate', 'Auditor', 'Bartender', 'Biologist', 'Bookkeeper', 'Buyer', 'Carpenter', 'Cashier', 'CEO', 'Clerk', 'Co-op', 'Co-Founder', 'Consultant', 'Coordinator', 'CTO', 'Developer', 'Designer', 'Director', 'Driver', 'Editor', 'Electrician', 'Engineer', 'Extern', 'Founder', 'Freelancer', 'Head', 'Intern', 'Janitor', 'Journalist', 'Laborer', 'Lawyer', 'Lead', 'Manager', 'Mechanic', 'Member', 'Nurse', 'Officer', 'Operator', 'Operation', 'Photographer', 'President', 'Producer', 'Recruiter', 'Representative', 'Researcher', 'Sales', 'Server', 'Scientist', 'Specialist', 'Supervisor', 'Teacher', 'Technician', 'Trader', 'Trainee', 'Treasurer', 'Tutor', 'Vice', 'VP', 'Volunteer', 'Webmaster', 'Worker'];\r\n\r\nconst hasJobTitle = (item: TextItem) =>\r\n  JOB_TITLES.some((jobTitle) =>\r\n    item.text.split(/\\s/).some((word) => word === jobTitle)\r\n  );\r\nconst hasMoreThan5Words = (item: TextItem) => item.text.split(/\\s/).length > 5;\r\nconst JOB_TITLE_FEATURE_SET: FeatureSet[] = [\r\n  [hasJobTitle, 4],\r\n  [hasNumber, -4],\r\n  [hasMoreThan5Words, -2],\r\n];\r\n\r\nexport const extractWorkExperience = (sections: ResumeSectionToLines) => {\r\n  const workExperiences: ResumeWorkExperience[] = [];\r\n  const workExperiencesScores = [];\r\n  const lines = getSectionLinesByKeywords(\r\n    sections,\r\n    WORK_EXPERIENCE_KEYWORDS_LOWERCASE\r\n  );\r\n  const subsections = divideSectionIntoSubsections(lines);\r\n\r\n  for (const subsectionLines of subsections) {\r\n    const descriptionsLineIdx = getDescriptionsLineIdx(subsectionLines) ?? 2;\r\n\r\n    const subsectionInfoTextItems = subsectionLines\r\n      .slice(0, descriptionsLineIdx)\r\n      .flat();\r\n    const [date, dateScores] = getTextWithHighestFeatureScore(\r\n      subsectionInfoTextItems,\r\n      DATE_FEATURE_SETS\r\n    );\r\n    const [jobTitle, jobTitleScores] = getTextWithHighestFeatureScore(\r\n      subsectionInfoTextItems,\r\n      JOB_TITLE_FEATURE_SET\r\n    );\r\n    const COMPANY_FEATURE_SET: FeatureSet[] = [\r\n      [isBold, 2],\r\n      [getHasText(date), -4],\r\n      [getHasText(jobTitle), -4],\r\n    ];\r\n    const [company, companyScores] = getTextWithHighestFeatureScore(\r\n      subsectionInfoTextItems,\r\n      COMPANY_FEATURE_SET,\r\n      false\r\n    );\r\n\r\n    const subsectionDescriptionsLines =\r\n      subsectionLines.slice(descriptionsLineIdx);\r\n    const descriptions = getBulletPointsFromLines(subsectionDescriptionsLines);\r\n\r\n    workExperiences.push({ company, jobTitle, date, descriptions });\r\n    workExperiencesScores.push({\r\n      companyScores,\r\n      jobTitleScores,\r\n      dateScores,\r\n    });\r\n  }\r\n  return { workExperiences, workExperiencesScores };\r\n};\r\n","import type { ResumeProject } from \"../../../lib/redux/types\";\r\nimport type {\r\n  FeatureSet,\r\n  ResumeSectionToLines,\r\n} from \"../../../lib/parse-resume-from-pdf/types\";\r\nimport { getSectionLinesByKeywords } from \"../../../lib/parse-resume-from-pdf/extract-resume-from-sections/lib/get-section-lines\";\r\nimport {\r\n  DATE_FEATURE_SETS,\r\n  getHasText,\r\n  isBold,\r\n} from \"../../../lib/parse-resume-from-pdf/extract-resume-from-sections/lib/common-features\";\r\nimport { divideSectionIntoSubsections } from \"../../../lib/parse-resume-from-pdf/extract-resume-from-sections/lib/subsections\";\r\nimport { getTextWithHighestFeatureScore } from \"../../../lib/parse-resume-from-pdf/extract-resume-from-sections/lib/feature-scoring-system\";\r\nimport {\r\n  getBulletPointsFromLines,\r\n  getDescriptionsLineIdx,\r\n} from \"../../../lib/parse-resume-from-pdf/extract-resume-from-sections/lib/bullet-points\";\r\n\r\nexport const extractProject = (sections: ResumeSectionToLines) => {\r\n  const projects: ResumeProject[] = [];\r\n  const projectsScores = [];\r\n  const lines = getSectionLinesByKeywords(sections, [\"project\"]);\r\n  const subsections = divideSectionIntoSubsections(lines);\r\n\r\n  for (const subsectionLines of subsections) {\r\n    const descriptionsLineIdx = getDescriptionsLineIdx(subsectionLines) ?? 1;\r\n\r\n    const subsectionInfoTextItems = subsectionLines\r\n      .slice(0, descriptionsLineIdx)\r\n      .flat();\r\n    const [date, dateScores] = getTextWithHighestFeatureScore(\r\n      subsectionInfoTextItems,\r\n      DATE_FEATURE_SETS\r\n    );\r\n    const PROJECT_FEATURE_SET: FeatureSet[] = [\r\n      [isBold, 2],\r\n      [getHasText(date), -4],\r\n    ];\r\n    const [project, projectScores] = getTextWithHighestFeatureScore(\r\n      subsectionInfoTextItems,\r\n      PROJECT_FEATURE_SET,\r\n      false\r\n    );\r\n\r\n    const descriptionsLines = subsectionLines.slice(descriptionsLineIdx);\r\n    const descriptions = getBulletPointsFromLines(descriptionsLines);\r\n\r\n    projects.push({ project, date, descriptions });\r\n    projectsScores.push({\r\n      projectScores,\r\n      dateScores,\r\n    });\r\n  }\r\n  return { projects, projectsScores };\r\n};\r\n","/**\r\n * Server side object deep clone util using JSON serialization.\r\n * Not efficient for large objects but good enough for most use cases.\r\n *\r\n * Client side can simply use structuredClone.\r\n */\r\nexport const deepClone = <T extends { [key: string]: any }>(object: T) =>\r\n  JSON.parse(JSON.stringify(object)) as T;\r\n","import { createSlice, type PayloadAction } from \"@reduxjs/toolkit\";\r\nimport type { RootState } from \"../../lib/redux/store\";\r\nimport type {\r\n  FeaturedSkill,\r\n  Resume,\r\n  ResumeEducation,\r\n  ResumeProfile,\r\n  ResumeProject,\r\n  ResumeSkills,\r\n  ResumeWorkExperience,\r\n} from \"../../lib/redux/types\";\r\nimport type { ShowForm } from \"../../lib/redux/settingsSlice\";\r\n\r\nexport const initialProfile: ResumeProfile = {\r\n  name: \"\",\r\n  summary: \"\",\r\n  email: \"\",\r\n  phone: \"\",\r\n  location: \"\",\r\n  url: \"\",\r\n};\r\n\r\nexport const initialWorkExperience: ResumeWorkExperience = {\r\n  company: \"\",\r\n  jobTitle: \"\",\r\n  date: \"\",\r\n  descriptions: [],\r\n};\r\n\r\nexport const initialEducation: ResumeEducation = {\r\n  school: \"\",\r\n  degree: \"\",\r\n  gpa: \"\",\r\n  date: \"\",\r\n  descriptions: [],\r\n};\r\n\r\nexport const initialProject: ResumeProject = {\r\n  project: \"\",\r\n  date: \"\",\r\n  descriptions: [],\r\n};\r\n\r\nexport const initialFeaturedSkill: FeaturedSkill = { skill: \"\", rating: 4 };\r\nexport const initialFeaturedSkills: FeaturedSkill[] = Array(6).fill({\r\n  ...initialFeaturedSkill,\r\n});\r\nexport const initialSkills: ResumeSkills = {\r\n  featuredSkills: initialFeaturedSkills,\r\n  descriptions: [],\r\n};\r\n\r\nexport const initialCustom = {\r\n  descriptions: [],\r\n};\r\n\r\nexport const initialResumeState: Resume = {\r\n  profile: initialProfile,\r\n  workExperiences: [initialWorkExperience],\r\n  educations: [initialEducation],\r\n  projects: [initialProject],\r\n  skills: initialSkills,\r\n  custom: initialCustom,\r\n};\r\n\r\n// Keep the field & value type in sync with CreateHandleChangeArgsWithDescriptions (components\\ResumeForm\\types.ts)\r\nexport type CreateChangeActionWithDescriptions<T> = {\r\n  idx: number;\r\n} & (\r\n  | {\r\n      field: Exclude<keyof T, \"descriptions\">;\r\n      value: string;\r\n    }\r\n  | { field: \"descriptions\"; value: string[] }\r\n);\r\n\r\nexport const resumeSlice = createSlice({\r\n  name: \"resume\",\r\n  initialState: initialResumeState,\r\n  reducers: {\r\n    changeProfile: (\r\n      draft,\r\n      action: PayloadAction<{ field: keyof ResumeProfile; value: string }>\r\n    ) => {\r\n      const { field, value } = action.payload;\r\n      draft.profile[field] = value;\r\n    },\r\n    changeWorkExperiences: (\r\n      draft,\r\n      action: PayloadAction<\r\n        CreateChangeActionWithDescriptions<ResumeWorkExperience>\r\n      >\r\n    ) => {\r\n      const { idx, field, value } = action.payload;\r\n      const workExperience = draft.workExperiences[idx];\r\n      workExperience[field] = value as any;\r\n    },\r\n    changeEducations: (\r\n      draft,\r\n      action: PayloadAction<CreateChangeActionWithDescriptions<ResumeEducation>>\r\n    ) => {\r\n      const { idx, field, value } = action.payload;\r\n      const education = draft.educations[idx];\r\n      education[field] = value as any;\r\n    },\r\n    changeProjects: (\r\n      draft,\r\n      action: PayloadAction<CreateChangeActionWithDescriptions<ResumeProject>>\r\n    ) => {\r\n      const { idx, field, value } = action.payload;\r\n      const project = draft.projects[idx];\r\n      project[field] = value as any;\r\n    },\r\n    changeSkills: (\r\n      draft,\r\n      action: PayloadAction<\r\n        | { field: \"descriptions\"; value: string[] }\r\n        | {\r\n            field: \"featuredSkills\";\r\n            idx: number;\r\n            skill: string;\r\n            rating: number;\r\n          }\r\n      >\r\n    ) => {\r\n      const { field } = action.payload;\r\n      if (field === \"descriptions\") {\r\n        const { value } = action.payload;\r\n        draft.skills.descriptions = value;\r\n      } else {\r\n        const { idx, skill, rating } = action.payload;\r\n        const featuredSkill = draft.skills.featuredSkills[idx];\r\n        featuredSkill.skill = skill;\r\n        featuredSkill.rating = rating;\r\n      }\r\n    },\r\n    changeCustom: (\r\n      draft,\r\n      action: PayloadAction<{ field: \"descriptions\"; value: string[] }>\r\n    ) => {\r\n      const { value } = action.payload;\r\n      draft.custom.descriptions = value;\r\n    },\r\n    addSectionInForm: (draft, action: PayloadAction<{ form: ShowForm }>) => {\r\n      const { form } = action.payload;\r\n      switch (form) {\r\n        case \"workExperiences\": {\r\n          draft.workExperiences.push(structuredClone(initialWorkExperience));\r\n          return draft;\r\n        }\r\n        case \"educations\": {\r\n          draft.educations.push(structuredClone(initialEducation));\r\n          return draft;\r\n        }\r\n        case \"projects\": {\r\n          draft.projects.push(structuredClone(initialProject));\r\n          return draft;\r\n        }\r\n      }\r\n    },\r\n    moveSectionInForm: (\r\n      draft,\r\n      action: PayloadAction<{\r\n        form: ShowForm;\r\n        idx: number;\r\n        direction: \"up\" | \"down\";\r\n      }>\r\n    ) => {\r\n      const { form, idx, direction } = action.payload;\r\n      if (form !== \"skills\" && form !== \"custom\") {\r\n        if (\r\n          (idx === 0 && direction === \"up\") ||\r\n          (idx === draft[form].length - 1 && direction === \"down\")\r\n        ) {\r\n          return draft;\r\n        }\r\n\r\n        const section = draft[form][idx];\r\n        if (direction === \"up\") {\r\n          draft[form][idx] = draft[form][idx - 1];\r\n          draft[form][idx - 1] = section;\r\n        } else {\r\n          draft[form][idx] = draft[form][idx + 1];\r\n          draft[form][idx + 1] = section;\r\n        }\r\n      }\r\n    },\r\n    deleteSectionInFormByIdx: (\r\n      draft,\r\n      action: PayloadAction<{ form: ShowForm; idx: number }>\r\n    ) => {\r\n      const { form, idx } = action.payload;\r\n      if (form !== \"skills\" && form !== \"custom\") {\r\n        draft[form].splice(idx, 1);\r\n      }\r\n    },\r\n    setResume: (draft, action: PayloadAction<Resume>) => {\r\n      return action.payload;\r\n    },\r\n  },\r\n});\r\n\r\nexport const {\r\n  changeProfile,\r\n  changeWorkExperiences,\r\n  changeEducations,\r\n  changeProjects,\r\n  changeSkills,\r\n  changeCustom,\r\n  addSectionInForm,\r\n  moveSectionInForm,\r\n  deleteSectionInFormByIdx,\r\n  setResume,\r\n} = resumeSlice.actions;\r\n\r\nexport const selectResume = (state: RootState) => state.resume;\r\nexport const selectProfile = (state: RootState) => state.resume.profile;\r\nexport const selectWorkExperiences = (state: RootState) =>\r\n  state.resume.workExperiences;\r\nexport const selectEducations = (state: RootState) => state.resume.educations;\r\nexport const selectProjects = (state: RootState) => state.resume.projects;\r\nexport const selectSkills = (state: RootState) => state.resume.skills;\r\n