Function computeTextAndOffsets

src/notebook.ts:39–89

(maybeNotebook: MaybeNotebook<T>)

// Note: Assumes that all notebooks are Python.
export function computeTextAndOffsets<T>(maybeNotebook: MaybeNotebook<T>): TextAndOffsets {
  const textModels = maybeNotebook.textModels ?? [];
  const modelLanguage = maybeNotebook.getLanguage(maybeNotebook.currentTextModel, undefined);
  const modelIsMarkdown = modelLanguage === Language.MARKDOWN;
  const modelIsExpected = isAllowedLanguage(modelLanguage);
  const relevantDocumentTexts: string[] = [];
  let additionalUtf8ByteOffset = 0;
  let found = false;
  for (const [idx, previousModel] of textModels.entries()) {
    if (modelIsExpected && maybeNotebook.currentTextModel === previousModel) {
      // There is an offset for all previous cells and the \n\n spacing after each one.
      additionalUtf8ByteOffset =
        relevantDocumentTexts
          .map((el) => numCodeUnitsToNumUtf8Bytes(el))
          .reduce((a, b) => a + b, 0) +
        '\n\n'.length * relevantDocumentTexts.length;
      found = true;
    }
    const previousModelLanguage = maybeNotebook.getLanguage(previousModel, idx);
    if (modelIsExpected && !modelIsMarkdown) {
      // Don't use markdown in the Python prompt construction.
      // TODO(prem): Consider adding as comments.
      if (previousModelLanguage === Language.MARKDOWN) {
        continue;
      } else if (previousModelLanguage === modelLanguage) {
        relevantDocumentTexts.push(maybeNotebook.getText(previousModel));
      }
    } else if (modelIsMarkdown) {
      if (previousModelLanguage === Language.MARKDOWN) {
        relevantDocumentTexts.push(maybeNotebook.getText(previousModel));
      } else if (isAllowedLanguage(previousModelLanguage)) {
        relevantDocumentTexts.push(
          `\`\`\`${
            NOTEBOOK_LANGUAGES[previousModelLanguage as AllowedLanguages]
          }\n${maybeNotebook.getText(previousModel)}\n\`\`\``
        );
      }
    }
  }
  const currentModelText = maybeNotebook.getText(maybeNotebook.currentTextModel);
  const text = found ? relevantDocumentTexts.join('\n\n') : currentModelText;
  const utf8ByteOffset = numCodeUnitsToNumUtf8Bytes(
    currentModelText,
    maybeNotebook.utf16CodeUnitOffset
  );
  return {
    text,
    utf8ByteOffset,
    additionalUtf8ByteOffset,
  };
}