Add sparse support for final convert

core/src/Config.ts

@@ -1,8 +1,5 @@
 import ItemTransformer from './transformer/ItemTransformer';
 
-export default interface Config {
-  // See DocumentInitParameters from https://mozilla.github.io/pdf.js/api/draft/module-pdfjsLib.html#DocumentInitParameters
-  pdfjsParams?: object;
+export default interface TransformConfig {
   transformers?: ItemTransformer[];
-  // TODO keep pdfPages ?
 }

core/src/Debugger.ts

@@ -3,7 +3,7 @@ import ItemTransformer from './transformer/ItemTransformer';
 import TransformContext from './transformer/TransformContext';
 import StageResult, { initialStage } from './debug/StageResult';
 import ColumnAnnotation from './debug/ColumnAnnotation';
-import AnnotatedColumn from './debug/AnnotatedColumn';
+import type AnnotatedColumn from './debug/AnnotatedColumn';
 import { detectChanges } from './debug/detectChanges';
 import { asPages } from './debug/Page';
 import EvaluationTracker from './debug/EvaluationTracker';

core/src/convert.ts

@@ -0,0 +1,115 @@
+import Globals from './Globals';
+import Item from './Item';
+import ParseProgressReporter from './ParseProgressReporter';
+import ParseResult from './ParseResult';
+import PdfParser from './PdfParser';
+import type ProgressListenFunction from './ProgressListenFunction';
+import { assert } from './assert';
+import ItemTransformer from './transformer/ItemTransformer';
+import TransformContext from './transformer/TransformContext';
+
+export type PdfSource = string | Uint8Array | object;
+
+export interface Converter {
+  convert: (items: Item[]) => string;
+}
+
+interface Options {
+  debug: boolean;
+  progressListener: ProgressListenFunction;
+}
+
+const defaultOptions: Options = {
+  debug: false,
+  progressListener: () => {},
+};
+
+export interface TransformationResult {
+  convert(converter: Converter): string;
+}
+
+export async function parseAndTransform(
+  src: PdfSource,
+  parser: PdfParser,
+  transformers: ItemTransformer[],
+  progressListener: ProgressListenFunction,
+): Promise<TransformationResult> {
+  const parseResult = await parseAndVerifyTransformers(src, parser, transformers, progressListener);
+  const transformedItems = transform(parseResult, transformers);
+  return Promise.resolve({
+    convert: (converter: Converter) => converter.convert(transformedItems),
+  });
+}
+
+export async function convert(
+  src: PdfSource,
+  parser: PdfParser,
+  transformers: ItemTransformer[],
+  converter: Converter,
+  options: Options = defaultOptions,
+): Promise<string> {
+  // parse
+  const parseResult = await parser.parse(src, new ParseProgressReporter(options.progressListener));
+  verifyRequiredColumns(parseResult.schema, transformers);
+
+  // transform
+  let items = parseResult.items;
+  let globals = new Globals();
+  const context = new TransformContext(parseResult.fontMap, parseResult.pageViewports, globals);
+  transformers.forEach((transformer) => {
+    const result = transformer.transform(context, items);
+    globals = globals.withValues(result.globals);
+    items = result.items;
+  });
+
+  // convert
+  return converter.convert(items);
+}
+
+async function parseAndVerifyTransformers(
+  src: PdfSource,
+  parser: PdfParser,
+  transformers: ItemTransformer[],
+  progressListener: ProgressListenFunction,
+): Promise<ParseResult> {
+  return parser.parse(src, new ParseProgressReporter(progressListener)).then((parseResult) => {
+    verifyRequiredColumns(parseResult.schema, transformers);
+    return parseResult;
+  });
+}
+
+/**
+ * Goes through all transformer and makes sure each required column is available in its predecessor schema.
+ *
+ * @param inputSchema
+ * @param transformers
+ */
+function verifyRequiredColumns(inputSchema: string[], transformers: ItemTransformer[]) {
+  const schemas: string[][] = [inputSchema];
+  for (let idx = 0; idx < transformers.length; idx++) {
+    const transformer = transformers[idx];
+    const predecessorSchema = schemas[idx];
+    transformer.descriptor.requireColumns?.forEach((column) => {
+      assert(
+        predecessorSchema.includes(column),
+        `Input schema [${predecessorSchema.join(', ')}] for transformer '${
+          transformer.name
+        }' does not contain the required column '${column}'`,
+      );
+    });
+    const outputSchema = transformer.schemaTransformer(predecessorSchema);
+    schemas.push(outputSchema);
+  }
+}
+
+function transform(parseResult: ParseResult, transformers: ItemTransformer[]) {
+  let items = parseResult.items;
+  let globals = new Globals();
+  const context = new TransformContext(parseResult.fontMap, parseResult.pageViewports, globals);
+  transformers.forEach((transformer) => {
+    const result = transformer.transform(context, items);
+    globals = globals.withValues(result.globals);
+    items = result.items;
+  });
+  return items;
+}

core/src/convert/MarkdownConverter.ts

@@ -0,0 +1,24 @@
+import Item from 'src/Item';
+import { Converter } from 'src/convert';
+
+export default class MarkdownConverter implements Converter {
+  convert(items: Item[]) {
+    let content = '';
+
+    items.forEach((item) => {
+      const types = item.data['types'] || [];
+      const itemText = item.data['str'];
+      if (types.includes('H1')) {
+        content += '# ' + itemText + '\n';
+      } else if (types.includes('H2')) {
+        content += '## ' + itemText + '\n';
+      } else if (types.includes('H3')) {
+        content += '## ' + itemText + '\n';
+      } else {
+        content += itemText;
+      }
+      content += '\n';
+    });
+    return content;
+  }
+}

core/src/index.ts

@@ -1,4 +1,4 @@
-import Config from './Config';
+import TransformConfig from './Config';
 import type ProgressListenFunction from './ProgressListenFunction';
 import ParseProgressReporter from './ParseProgressReporter';
 import PdfParser from './PdfParser';
@@ -12,7 +12,9 @@ import CompactLines from './transformer/CompactLines';
 import SortXWithinLines from './transformer/SortXWithinLines';
 import RemoveRepetitiveItems from './transformer/RemoveRepetitiveItems';
 import DetectToc from './transformer/DetectToc';
+import DetectHeaders from './transformer/DetectHeaders';
 import NoOpTransformer from './transformer/NoOpTransformer';
+import {type ParseConfig } from './parse';
 
 export const transformers = [
   new AdjustHeight(),
@@ -23,27 +25,20 @@ export const transformers = [
   new SortXWithinLines(),
   new RemoveRepetitiveItems(),
   new DetectToc(),
+  new DetectHeaders(),
   new NoOpTransformer(),
 ];
 
-const defaultConfig: Config = {
-  pdfjsParams: {
-    // TODO check if that cmap thing makes sense since we don't bundle them
-    cMapUrl: 'cmaps/',
-    cMapPacked: true,
-  },
-  transformers,
-};
-
-export function pdfParser(pdfJs: any) {
-  return new PdfParser(pdfJs, defaultConfig.pdfjsParams);
+export interface Options {
+  parseConfig?: ParseConfig;
+  transformConfig?: TransformConfig;
 }
 
 export function parseReporter(progressListener: ProgressListenFunction) {
   return new ParseProgressReporter(progressListener);
 }
 
-export function createPipeline(pdfJs: any, config = defaultConfig): PdfPipeline {
-  const parser = new PdfParser(pdfJs, config.pdfjsParams);
-  return new PdfPipeline(parser, config.transformers || transformers);
+export function createPipeline(pdfJs: any, options: Options = {}): PdfPipeline {
+  const parser = new PdfParser(pdfJs);
+  return new PdfPipeline(parser, options.transformConfig?.transformers || transformers);
 }

core/src/parse.ts

@@ -0,0 +1,19 @@
+import PdfParser from './PdfParser';
+
+export interface ParseConfig {
+  // See DocumentInitParameters from https://mozilla.github.io/pdf.js/api/draft/module-pdfjsLib.html#DocumentInitParameters
+  pdfjsParams?: object;
+  // TODO keep pdfPages ?
+}
+
+const defaultConfig: ParseConfig = {
+  pdfjsParams: {
+    // TODO check if that cmap thing makes sense since we don't bundle them
+    cMapUrl: 'cmaps/',
+    cMapPacked: true,
+  },
+};
+
+export function parser(pdfJs: any, options: ParseConfig = defaultConfig) {
+  return new PdfParser(pdfJs, options.pdfjsParams);
+}