recipe parser

See change history

In its current incarnation, lexes, parses, and converts, lists of ingredients and steps into a JSON object (see below) used by the recipe-ui project. The goal of the project is to parse entire recipes.

This is work in progress.

A shout-out to the awesome libraries the recipe parser relies on

1. Chevrotain for lexing, parsing, and semantics.
2. pluralize for pluralization.
3. XRegExp for sanity using regular expressions.

The following code will parse a list of ingredients into a JSON object

const text = `dough
        1 1/2 cp all-purpose flour
        1 tsp vanilla extract,
        sauce
        1 cup milk
        1 egg`
const {recipe, errors} = toRecipe(text, {inputType: ParseType.INGREDIENTS})

...and the result...

{
  "type": "ingredients",
  "ingredients": [
    {
      "amount": {
        "quantity": 1.5,
        "unit": "cup"
      },
      "ingredient": "all-purpose flour",
      "section": "dough"
    },
    {
      "amount": {
        "quantity": 1,
        "unit": "tsp"
      },
      "ingredient": "vanilla extract",
      "section": "dough"
    },
    {
      "amount": {
        "quantity": 1,
        "unit": "cup"
      },
      "ingredient": "milk",
      "section": "sauce"
    },
    {
      "amount": {
        "quantity": 1,
        "unit": "piece"
      },
      "ingredient": "egg",
      "section": "sauce"
    }
  ]
}

As a more complete example, we parse the text for a fake Piri-Piri chicken recipe.

    it("should be able to parse the piri piri chicken recipe", () => {
    const input = `Ingredients
        Powder
            1. 2 tbsp sugar
            2) 1 tbsp paprika
            1 tbsp coriander
            1 tbsp cumin
            1 1/2 tbsp salt
            2 tbsps new mexico chile powder
            Sauce
            3 cloves garlic
            8 fresno peppers
            1/3 cup lemon juice
            1/4 cup red wine vinegar
            Chicken
            1 whole chicken
            Steps
            Sauce
            1. first step
            2. second step
            Chicken
            3) third step
            `
    const {recipe, errors} = toRecipe(input, {deDupSections: true})
    expect(recipe).toEqual({
        type: "recipe",
        ingredients: [
            {amount: {quantity: 2, unit: UnitType.TABLESPOON}, ingredient: 'sugar', section: 'Powder', brand: null},
            {amount: {quantity: 1, unit: UnitType.TABLESPOON}, ingredient: 'paprika', section: null, brand: null},
            {amount: {quantity: 1, unit: UnitType.TABLESPOON}, ingredient: 'coriander', section: null, brand: null},
            {amount: {quantity: 1, unit: UnitType.TABLESPOON}, ingredient: 'cumin', section: null, brand: null},
            {amount: {quantity: 1.5, unit: UnitType.TABLESPOON}, ingredient: 'salt', section: null, brand: null},
            {amount: {quantity: 2, unit: UnitType.TABLESPOON}, ingredient: 'new mexico chile powder', section: null, brand: null},

            {amount: {quantity: 3, unit: UnitType.PIECE}, ingredient: 'cloves garlic', section: 'Sauce', brand: null},
            {amount: {quantity: 8, unit: UnitType.PIECE}, ingredient: 'fresno peppers', section: null, brand: null},
            {amount: {quantity: 0.3333333333333333, unit: UnitType.CUP}, ingredient: 'lemon juice', section: null, brand: null},
            {amount: {quantity: 0.25, unit: UnitType.CUP}, ingredient: 'red wine vinegar', section: null, brand: null},

            {amount: {quantity: 1, unit: UnitType.PIECE}, ingredient: 'whole chicken', section: 'Chicken', brand: null},
        ],
        steps: [
            {id: "1.", step: "first step", title: "Sauce"},
            {id: "2.", step: "second step", title: null},
            {id: "3)", step: "third step", title: "Chicken"},
        ]
    })
    expect(errors).toHaveLength(0)
})

usage

To use the recipe parser, add the library to your project

npm install @saucie/recipe-parser

and install the peer dependencies

npm install chevrotain
npm install pluralize
npm install xregexp

Version compatibility

package	support versions	notes
chevrotain	>= 10.0.0	9.0.0 works but the changes the CST nodes to be possibly undefined (typescript)
pluralize	>= 7.0.0	Possibly works with lower versions (let me know)
xregexp	>= 4.0.0	Possibly works with lower versions (let me know)

Add an import to your module

import {toRecipe} from "@saucie/recipe-parser";

And then call the toRecipe(...) function with any options

import {toRecipe, ParseType} from "@saucie/recipe-parser";

const myRecipe = "some recipe text"
const {recipe, errors} = toRecipe(myRecipe, {deDupSections: true, inputType: ParseType.RECIPE})

If there are no errors, the recipe will be a JSON object of type Recipe

export type Recipe = {
    type: string
    ingredients: Array<Ingredient>
    steps: Array<Step>
}

// which depends on the following definitions
export type Ingredient = {
    amount: Amount
    ingredient: string
    section: string | null
    brand: string | null
}

export type Step = {
    id: string
    title: string | null
    step: string
}

export type Amount = {
    quantity: number
    unit: Unit
}

enum Unit {
    MILLIGRAM = 'mg', GRAM = 'g', KILOGRAM = 'kg',
    OUNCE = 'oz', POUND = 'lb',
    MILLILITER = 'ml', LITER = 'l', TEASPOON = 'tsp', TABLESPOON = 'tbsp', FLUID_OUNCE = 'fl oz',
    CUP = 'cup', PINT = 'pt', QUART = 'qt', GALLON = 'gal',
    PIECE = 'piece', PINCH = 'pinch'
}

/**
 * Converts the text to a list of recipe ingredients with optional sections. This is the
 * function to call to convert a test recipe into a recipe object.
 * @param text The text to convert into a recipe object
 * @param [options = defaultOptions] The options used for parsing the text into a
 * recipe or recipe fragment.
 * @return A recipe result holding the recipe object and any parsing errors
 */
function toRecipe(text: string, options: Options = defaultOptions): RecipeResult {/*...*/}

and the options are

export type Options = {
    // When set to `true` only sets the section of the first ingredient of each
    // section to current section.
    deDupSections?: boolean
    // When set to `true` then logs warning to the console, otherwise
    // does not log warnings. Warning and errors are reported in the returned object
    // in either case.
    logWarnings?: boolean
    // The thing that the input text represents: a whole recipe, a list of ingredients,
    // or a list of steps.
    inputType?: ParseType
}

and the parse-type is defined as

export enum ParseType {
    RECIPE,
    INGREDIENTS,
    STEPS
}

and the recipe result is defined as

/**
 * The result of the lexing, parsing, and visiting.
 */
export type RecipeResult = {
    recipe: Recipe,
    errors: Array<ILexingError>
}

the grammar

In order for a recipe's ingredients to be parsed, then must adhere to the following grammar, which uses the Augmented Backus-Naur Form (ABNF) notation (see this nice article) for an introduction to grammar notations.

// an ingredient list has either sections or ingredient itmes or both
ingredients = *[section] *[ingredient_item]

// a section has a header and a list of ingredient items
section = section_header 1*ingredient_item

// a section header must be on its own line, or surrounded by "#" or some
// combination of thos
section_header = (newline / "#") 1*word (*["#"] / newline)

// an ingredient item has an optional list ID, an amount, and an ingredient
ingredient_item = [ingredient_item_id 1*whitespace] amount ingredient

// an ingredient item ID is a list item, for example "1.", "*", "-", "•", "1)", etc
ingredient_item_id = ( [ "(" ] number [ "." / ")" / ":" ] ) / ( [ "-" / "*" / "•" ])

// the amount is a quantity and an optional unit (when no unit is present, will be
// treated as pieces, such as "1 egg" would become "1 piece egg")
amount = quantity [white_space] [unit]

// the ingredient (e.g. egg, all-purpose flour, etc) is a sequence of words that
// ends in a newline
ingredient = *word newline

// the quantity is a number or a fraction. A fraction can be expressed as a whole
// number and a fractional part (e.g. 1 1/4) or as a whole number and a unicode
// fraction
quantity = number / fraction

// units can be abbreviated, can be synonyms, plural...see the "unitMatcher(..)" function
// in the RecipeLexer.ts file for more details
unit = (mg / g / kg / oz / lb / ml / l / tsp / tbsp/ fl oz / cup / pt / qt / gallon /)["."]

// a word
word = 1*("\w" / "." / "'" / "(" / ")" / "[" / "]" / "{" / "}" / "-")

number = integer / decimal / (integer unicode_fraction)
integer = 0 / (natural_digit *digit)
decimal = integer "." 1*digit
fraction = integer "/" natural_digit *digit
natural_digit = 1 / 2 / 3 / 4 / 5 / 6 / 7 / 8 / 9
digit = 0 / natural_digit
unicode_fraction = \u00BC | \u00BD | \u00BE | ...

newline = "\n" / "\r\n"
white_space = *( " " / "\t" )

Recipes using this grammar get parsed into the following syntax tree structure.