ParserController.scala 10 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293
  1. package com.weEat.controllers
  2. import com.weEat.shared.models._
  3. import javax.inject.{Inject,Singleton}
  4. import play.api.libs.json._
  5. import play.api.mvc._
  6. import scala.concurrent.Future
  7. import com.weEat.models.Authorization
  8. import scalaoauth2.provider.{AuthInfoRequest,OAuth2ProviderActionBuilders}
  9. import com.weEat.services.OAuth2Service
  10. import net.ruippeixotog.scalascraper.browser.JsoupBrowser
  11. import net.ruippeixotog.scalascraper.dsl.DSL._
  12. import net.ruippeixotog.scalascraper.dsl.DSL.Extract._
  13. //import net.ruippeixotog.scalascraper.dsl.DSL.Parse._
  14. import net.ruippeixotog.scalascraper.model.Element
  15. import net.ruippeixotog.scalascraper.scraper.HtmlExtractor
  16. import scala.util._
  17. @Singleton
  18. class ParserController @Inject()(
  19. val controllerComponents: ControllerComponents,
  20. oauth: OAuth2Service,
  21. usdaController: USDAController,
  22. foodController: FoodController
  23. ) extends BaseController
  24. with OAuth2ProviderActionBuilders {
  25. implicit val ec = scala.concurrent.ExecutionContext.global
  26. private val _browser = JsoupBrowser()
  27. def parseURL() = AuthorizedAction[Authorization](oauth).async(parse.text)({ implicit request: AuthInfoRequest[String, Authorization] =>
  28. val url = request.body
  29. _findParser(url).fold(
  30. Future.successful(NotFound(s"No parser available for $url."))
  31. ) { (parser) =>
  32. val doc = _browser.get(url)
  33. val title = doc >> parser.titleExtractor
  34. val servings = doc >> parser.servingExtractor
  35. val prepTime = parser.prepTimeExtractor.flatMap(doc >?> _)
  36. val cookTime = parser.cookTimeExtractor.flatMap(doc >?> _)
  37. val ingredients = doc >> parser.ingredientExtractor
  38. val instructions = doc >> parser.instructionExtractor
  39. Future.sequence(ingredients.map({
  40. case (amt, u, line) => _guessFoodFromStr(line).map(Ingredient(_, amt, u))
  41. }))
  42. .map((ingredients) => Ok(Json.toJson(RecipeNodeNoId(
  43. title,
  44. servings.getOrElse(1.0f),
  45. 1.0f,
  46. UnitType.NUMBER,
  47. ingredients.toSeq,
  48. /* tflucke@[2023-10-26]: Do not pss along the instructions since this
  49. * could be a violation of the Recipe Author's copyright. */
  50. Nil, //instructions.toSeq,
  51. None,
  52. None,
  53. Some(url),
  54. None
  55. ))))
  56. }
  57. })
  58. private def _findParser(url: String): Option[Parser] = {
  59. val host = new java.net.URL(url).getAuthority()
  60. val hostNoWWW =
  61. if (host.startsWith("www.")) host.substring("www.".length) else host
  62. Map(
  63. ("epicurious.com" -> Parser.epicurious),
  64. ("mccormick.com" -> Parser.mccormick),
  65. ("recipetineats.com" -> Parser.recipeTinEats),
  66. ("mamalovestocook.com" -> Parser.recipeTinEats),
  67. ("soulfullymade.com" -> Parser.recipeTinEats),
  68. ("sallysbakingaddiction.com" -> Parser.sallysBakingAddiction),
  69. ("seriouseats.com" -> Parser.seriousEats),
  70. ("greatist.com" -> Parser.greatist),
  71. ).get(hostNoWWW)
  72. }
  73. private def _guessFoodFromStr(
  74. foodLine: String
  75. ): Future[Ingredient.IngredientId] = {
  76. import gov.usda.nal.fdc.models.DataType._
  77. usdaController.fdc.getFoodsSearch(foodLine
  78. .filter(_ <= 0x7f)
  79. .filterNot(_ == ':')
  80. .filterNot(_ == '/'), Seq(
  81. // Branded,
  82. Foundation, SRLegacy
  83. ), pageSize = Some(10))().flatMap({ (fdcResult) =>
  84. Future.sequence(
  85. fdcResult.foods.map((food) => foodController.getByFdcId(food.fdcId))
  86. ).map(_.flatten
  87. .headOption
  88. .fold[Ingredient.IngredientId](
  89. Ingredient.USDAId(fdcResult.foods.head.fdcId)
  90. )((foodNode) => Ingredient.FoodNodeId(foodNode._id))
  91. ).transform({
  92. case Success(x) => Success(x)
  93. case Failure(x) => println(s"Food lookup failed: $x");Failure(x)
  94. })
  95. })
  96. }
  97. }
  98. case class Parser(
  99. titleExtractor: HtmlExtractor[Element, String],
  100. servingExtractor: HtmlExtractor[Element, Option[Float]],
  101. prepTimeExtractor: Option[HtmlExtractor[Element, String]],
  102. cookTimeExtractor: Option[HtmlExtractor[Element, String]],
  103. ingredientExtractor: HtmlExtractor[Element, Iterable[(Float, MeasureUnit, String)]],
  104. instructionExtractor: HtmlExtractor[Element, Iterable[String]],
  105. )
  106. object Parser {
  107. val mccormick = Parser(
  108. text("h1"),
  109. // TODO use extractors
  110. text(".main-title .count").map(_.toFloatOption),
  111. Some(text(".prep_time .first_content")),
  112. cookTimeExtractor = Some(text(".ingredients .first_content")),
  113. ingredientExtractor = texts(".recipe-about-list li").map(
  114. _.map(_parseIngredient _)
  115. ),
  116. texts(".instructions-main span.para")
  117. )
  118. val epicurious = Parser(
  119. text("h1"),
  120. text("""div[data-testid="IngredientList"] > p""")
  121. .map("Yield: \\D*(\\d+).*".r.findFirstMatchIn(_).map(_.group(1).toFloat)),
  122. None,
  123. None,
  124. texts("""div[data-testid="IngredientList"] > div > div""").map(
  125. _.map(_parseIngredient _)
  126. ),
  127. texts("""div[data-testid="InstructionsWrapper"] > ol > li > p""")
  128. )
  129. val recipeTinEats = Parser(
  130. text("h2.wprm-recipe-name"),
  131. text("span.wprm-recipe-servings").map(_.toFloatOption),
  132. Some(text("span.wprm-recipe-prep_time-minutes")),
  133. Some(text("span.wprm-recipe-cook_time-minutes")),
  134. texts("li.wprm-recipe-ingredient")
  135. .map(_.map(_
  136. .replaceAll("\u00BD", "1/2")
  137. .replaceAll("\u00BC", "1/4")
  138. .replaceAll("\u00BE", "3/4")
  139. .replaceAll("\u2150", "1/7")
  140. .replaceAll("\u2151", "1/9")
  141. .replaceAll("\u2152", "1/10")
  142. .replaceAll("\u2153", "1/3")
  143. .replaceAll("\u2154", "2/3")
  144. .replaceAll("\u2155", "1/5")
  145. .replaceAll("\u2156", "2/5")
  146. .replaceAll("\u2157", "3/5")
  147. .replaceAll("\u2158", "4/5")
  148. .replaceAll("\u2159", "1/6")
  149. .replaceAll("\u215A", "5/6")
  150. .replaceAll("\u215B", "1/8")
  151. .replaceAll("\u215C", "3/8")
  152. .replaceAll("\u215D", "5/8")
  153. .replaceAll("\u215E", "7/8")
  154. .replaceAll("\u215F", "1/")
  155. .replaceAll("\u00F1", "n")
  156. .trim
  157. ))
  158. .map(_.map(_parseIngredient _)),
  159. texts("div.wprm-recipe-instruction-text")
  160. )
  161. val sallysBakingAddiction = Parser(
  162. text("h2.tasty-recipes-title"),
  163. text("span.tasty-recipes-yield")
  164. .map("\\D*(\\d+).*".r.findFirstMatchIn(_).map(_.group(1).toFloat)),
  165. Some(text("span.tasty-recipes-prep-time")),
  166. Some(text("span.tasty-recipes-cook-time")),
  167. elementList("div.tasty-recipes-ingredients-body > ul > li").map(
  168. _.map({(listItem) => (
  169. ((listItem >?> elementList("span"))
  170. .map(_.last)
  171. .fold(0.0f)((elm: Element) =>
  172. (elm >?> attr("data-amount"))
  173. .fold(0.0f)(_.toFloat)
  174. )
  175. ),
  176. (listItem >?> elementList("span"))
  177. .map(_.last)
  178. .fold[MeasureUnit](Gram)((elm: Element) =>
  179. (elm >?> attr("data-unit"))
  180. .flatMap(MeasureUnit.guessUnit _)
  181. .getOrElse(Count)
  182. ),
  183. listItem >> text("strong")
  184. )})
  185. ),
  186. texts("div.tasty-recipes-instructions-body > ol > li")
  187. )
  188. val seriousEats = Parser(
  189. text("h2.recipe-decision-block__title"),
  190. text("div.recipe-serving > span > span.meta-text__data")
  191. .map("\\D*(\\d+).*".r.findFirstMatchIn(_).map(_.group(1).toFloat)),
  192. //text("div.recipe-yield > span > span.meta-text__data")
  193. Some(text("div.prep-time > span > span.meta-text__data")),
  194. None, //Some(text("span.tasty-recipes-cook-time")),
  195. elementList("ul.structured-ingredients__list > li > p").map(
  196. _.map({(p) => (
  197. ((p >?> elementList("span"))
  198. .flatMap(_
  199. .filter((s) => (s >?> attr("data-ingredient-quantity")).isDefined)
  200. .lastOption
  201. .map(_ >> text)
  202. ).flatMap(_parseFraction _)
  203. .getOrElse(0.0f)
  204. ),
  205. ((p >?> elementList("span"))
  206. .flatMap(_
  207. .filter((s) => (s >?> attr("data-ingredient-unit")).isDefined)
  208. .lastOption
  209. .map(_ >> text)
  210. ).flatMap(MeasureUnit.guessUnit _)
  211. .getOrElse(Count)
  212. ),
  213. ((p >?> elementList("span"))
  214. .flatMap(_
  215. .filter((s) => (s >?> attr("data-ingredient-name")).isDefined)
  216. .headOption
  217. ).getOrElse(p) >> text
  218. )
  219. )})
  220. ),
  221. texts("div.structured-project__steps_1-0 > ol > li > p")
  222. )
  223. val greatist = Parser(
  224. text("h1"),
  225. elementList("article.article-body > ul > li").map(
  226. _.filter((listItem) => (listItem >?> text("strong")) == Some("Yield"))
  227. .map(_ >> text)
  228. .head
  229. ).map("Yield: \\D*(\\d+).*".r.findFirstMatchIn(_).map(_.group(1).toFloat)),
  230. // tflucke@[2023-11-28]: TODO They don't give passive, only Active + Total
  231. None,
  232. Some(
  233. elementList("article.article-body > ul > li").map(
  234. _.filter((listItem) => (listItem >?> text("strong")) == Some("Active"))
  235. .map(_ >> text)
  236. .head
  237. ).map("Active: \\D*(\\d+).*".r.findFirstMatchIn(_).fold("0")(_.group(1)))
  238. ),
  239. elementList("article.article-body > ul > li").map(_
  240. .filter((listItem) => (listItem >?> text("strong")) == None)
  241. .map(_ >> text)
  242. .map(_.replaceAll("\u00F1", "n"))
  243. .map(_parseIngredient _)
  244. ),
  245. texts("article.article-body > ol > li")
  246. )
  247. private def _parseFraction(fractionLine: String) = {
  248. val fractionPattern = raw"(\d+)/(\d+)".r
  249. val mixedFractionPattern = raw"(\d+)\w+(\d+)/(\d+)".r
  250. fractionLine match {
  251. case fractionPattern(numerator, denominator) =>
  252. Some(numerator.toFloat/denominator.toFloat)
  253. case mixedFractionPattern(whole, numerator, denominator) =>
  254. Some(whole.toFloat + numerator.toFloat/denominator.toFloat)
  255. case _ => fractionLine.toFloatOption
  256. }
  257. }
  258. private def _parseIngredient(
  259. ingredientLine: String
  260. ): (Float, MeasureUnit, String) = {
  261. val numberPattern = raw"(\d+)[\d-_]*\s(\w+)\s+(.+)".r
  262. val fractionPattern = raw"(\d+)/(\d+)[\d-_]*\s(\w+)\s+(.+)".r
  263. ingredientLine match {
  264. case numberPattern(amount, unit, rest) =>
  265. (amount.toFloat, MeasureUnit.guessUnit(unit).getOrElse(Count), rest)
  266. case fractionPattern(numerator, denominator, unit, rest) =>
  267. (
  268. numerator.toFloat/denominator.toFloat,
  269. MeasureUnit.guessUnit(unit).getOrElse(Count),
  270. rest
  271. )
  272. case noUnitLine =>
  273. (1, Count, noUnitLine)
  274. }
  275. }
  276. }