Fixed imports of PDF text with superscripts; added new test

scribeocr · Nov 9, 2024 · add376e · add376e
1 parent 9361df4
commit add376e
Show file tree

Hide file tree

Showing 3 changed files with 22 additions and 2 deletions.
diff --git a/js/import/convertPageStext.js b/js/import/convertPageStext.js
@@ -82,6 +82,7 @@ export async function convertPageStext({ ocrStr, n }) {
       const bboxes = [];
 
       const baselineSlopeArr = /** @type {Array<Number>} */ ([]);
+      let baselineFirstDone = false;
       const baselineFirst = /** @type {Array<Number>} */ ([]);
 
       let baselineCurrent = 0;
@@ -193,16 +194,21 @@ export async function convertPageStext({ ocrStr, n }) {
                 bboxesWordArr = [];
               }
 
-              // If the first word was determined to be a superscript, reset `baselineFirst` to avoid skewing the slope calculation.
               if (sizeDelta > 0) {
-                baselineFirst.length = 0;
+                // If the first word was determined to be a superscript, reset `baselineFirst` to avoid skewing the slope calculation.
+                if (!baselineFirstDone) baselineFirst.length = 0;
                 familyCurrent = fontNameStrI || familyCurrent;
                 sizeCurrent = sizeCurrentRaw || sizeCurrent;
                 fontSizeWord = sizeCurrent;
                 fontFamily = familyCurrent;
                 superArr[superArr.length - 1] = true;
               }
 
+              // If `baselineFirstDone` was set using a non-superscript word, mark it as done.
+              if (superArr.length > 0 && !superArr[superArr.length - 1] && baselineFirst.length > 0) {
+                baselineFirstDone = true;
+              }
+
               superCurrent = sizeDelta < 0;
             } else {
               sizeCurrent = sizeCurrentRaw || sizeCurrent;

diff --git a/tests/assets/superscript_examples_rotated.pdf b/tests/assets/superscript_examples_rotated.pdf
diff --git a/tests/module/importPdfText.spec.js b/tests/module/importPdfText.spec.js
@@ -247,3 +247,17 @@ describe('Check that PDF imports split lines correctly.', function () {
     await scribe.terminate();
   });
 }).timeout(120000);
+
+describe('Check that line baselines are imported correctly.', function () {
+  this.timeout(10000);
+
+  it('Should correctly parse line baselines for pages with rotation', async () => {
+    await scribe.importFiles([`${ASSETS_PATH_KARMA}/superscript_examples_rotated.pdf`], { extractPDFTextNative: true, extractPDFTextOCR: true });
+    assert.strictEqual(Math.round(scribe.data.ocr.active[0].lines[25].baseline[1]), -10);
+    assert.strictEqual(Math.round(scribe.data.ocr.active[1].lines[25].baseline[1]), -165);
+  }).timeout(10000);
+
+  after(async () => {
+    await scribe.terminate();
+  });
+}).timeout(120000);