chrome/browser/lens/core/mojom/text.mojom

// Copyright 2024 The Chromium Authors
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.

// These Mojo structs are converted from
// third_party/lens_server_proto/lens_overlay_text.proto and
// third_party/lens_server_proto/lens_overlay_deep_gleam_data.proto.
module lens.mojom;

import "chrome/browser/lens/core/mojom/geometry.mojom";
import "ui/gfx/geometry/mojom/geometry.mojom";
import "mojo/public/mojom/base/big_buffer.mojom";
import "skia/public/mojom/skcolor.mojom";

// The text reading order.
enum WritingDirection {
  kLeftToRight = 0,
  kRightToLeft = 1,
  kTopToBottom = 2,
};

// The text alignment.
enum Alignment {
  kDefaultLeftAlgined = 0,
  kRightAligned = 1,
  kCenterAligned = 2,
};

struct Text {
  // Information regarding the layout and position of text.
  TextLayout? text_layout;

  // Dominant content language of the text. Language code is CLDR/BCP-47.
  string? content_language;
};

// Nested text structure.
struct TextLayout {
  // The actual detected text on the page in an array of paragraphs and the
  // translated text if provided.
  array<Paragraph> paragraphs;
};

struct Word {
  // Required. The text in a plain text, which can be an empty string.
  string plain_text;

  // Optional. The text separator that should be appended after this word when
  // it is concatenated with the subsequent word in the same or next
  // line/paragraph into a single-line string.
  string? text_separator;

  // Optional. The geometry of the word.
  Geometry? geometry;

  // Optional. The text writing direction (aka reading order). All words in a
  // paragraph must have the same writing direction.
  WritingDirection? writing_direction;

  // Optional. Metadata for formulas. This is populated for entities of
  // `type=FORMULA`.
  FormulaMetadata? formula_metadata;
};

struct Line {
  // Optional. List of words in natural reading order.
  array<Word> words;

  // Optional. The geometry of the line.
  Geometry? geometry;
};

struct BackgroundImageData {
  // Image bytes to inpaint the source text.
  mojo_base.mojom.BigBuffer background_image;

  // Size of background_image in pixels.
  gfx.mojom.Size image_size;

  // Vertical padding to apply to the text box before drawing the background
  // image. Expressed as a fraction of the text box height, i.e. 1.0 means
  // that the height should be doubled. Half of the padding should be added on
  // the top and half on the bottom.
  float vertical_padding;

  // Horizontal padding to apply to the text box before drawing the background
  // image. Expressed as a fraction of the text box height. Half of the
  // padding should be added on the left and half on the right.
  float horizontal_padding;

  // Text mask for the generated background image.
  mojo_base.mojom.BigBuffer text_mask;
};

struct TranslatedLine {
  // Optional. List of words in natural reading order.
  array<Word> words;

  // Full translation of the line.
  string translation;

  // Required. The color of the text to render.
  skia.mojom.SkColor text_color;

  // Required. The color of the background used to mask the actual text.
  skia.mojom.SkColor background_primary_color;

  // Optional. The background image data used to mask the text underneath the
  // line.
  BackgroundImageData? background_image_data;

  // Optional. The geometry of the line.
  Geometry? geometry;
};

// This struct is used to transport a response from a trusted server to be
// displayed/consumed in WebUI. Specifically, any detected text that was
// requested to be translated in the server request.
struct TranslatedParagraph {
  // Optional. List of lines in natural reading order (see also
  // `writing_direction`). The amount of translated lines should be equal to the
  // amount of lines returned in the detected text. See comment in `Paragraph`
  // struct.
  array<TranslatedLine> lines;

  // Size of the resized screenshot in pixels.
  gfx.mojom.Size resized_bitmap_size;

  // Optional. The alignment of the text in the paragraph.
  Alignment? alignment;

  // Optional. The text writing direction (aka reading order).
  WritingDirection? writing_direction;

  // Optional. BCP-47 language code of the dominant language in this
  // paragraph.
  string? content_language;
};

// This struct is used to transport a response from a trusted server to be
// displayed/consumed in WebUI.  Specifically, any detected text that was found
// on the full page screenshot sent in the original server request.
struct Paragraph {
  // Optional. List of lines in natural reading order (see also
  // `writing_direction`).
  array<Line> lines;

  // Optional. A paragraph struct corresponding to this one containing the
  // translated text if provided. When present, the number of elements in
  // translation.lines must match the number of elements in lines. This is
  // normally discouraged in Mojo, but there isn't a good way to
  // represent/enforce this sort of invariant purely through struct definitions.
  // This is considered acceptable here, since this data only flows from a
  // trusted source to an untrusted sink.
  TranslatedParagraph? translation;

  // Optional. Geometry of the paragraph.
  Geometry? geometry;

  // Optional. The text writing direction (aka reading order).
  WritingDirection? writing_direction;

  // Optional. BCP-47 language code of the dominant language in this
  // paragraph.
  string? content_language;
};

struct FormulaMetadata {
  // Optional. LaTeX representation of a formula. Can be the same as
  // `plain_text`. Example: "\frac{2}{x}=y".
  string? latex;
};