languagetool-org/languagetool · error · RuntimeException

Check failed on sentence: ${StringUtils.abbreviate(sentence.

Error message

Check failed on sentence: ${StringUtils.abbreviate(sentence.getText(), 250)}

What it means

During SentenceSourceChecker.run, if checking an individual sentence throws any exception other than the control-flow DocumentLimitReachedException/ErrorLimitReachedException, it is wrapped in a RuntimeException quoting the first 250 characters of the sentence. Unless --skip-exceptions is set, one bad sentence aborts the whole run. The wrap exists to pinpoint which sentence broke the check.

Source

Thrown at languagetool-wikipedia/src/main/java/org/languagetool/dev/dumpcheck/SentenceSourceChecker.java:278

          if (options.hasOption("print-correct")) {
            if (matches.getRuleMatches().size() == 0) {
              System.out.println(sentence.getText());
            }
          } else {
            resultHandler.handleResult(sentence, matches.getRuleMatches(), lang);
          }
          sentenceCount++;
          if (sentenceCount % 5000 == 0) {
            System.err.printf("%s sentences checked...\n", NumberFormat.getNumberInstance(Locale.US).format(sentenceCount));
          }
          ruleMatchCount += matches.getRuleMatches().size();
        } catch (DocumentLimitReachedException | ErrorLimitReachedException e) {
          throw e;
        } catch (Exception e) {
          if (options.hasOption("skip-exceptions")) {
            e.printStackTrace();
          } else {
            throw new RuntimeException("Check failed on sentence: " + StringUtils.abbreviate(sentence.getText(), 250), e);
          }
        }
      }
      ignoredCount = mixingSource.getIgnoredCount();
    } catch (DocumentLimitReachedException | ErrorLimitReachedException e) {
      System.out.println(getClass().getSimpleName() + ": " + e);
    } finally {
      lt.shutdown();
      if (resultHandler != null) {
        System.out.printf(lang + ": %d total matches\n", ruleMatchCount);
        System.out.printf(lang + ": %d total sentences considered\n", sentenceCount);
        float matchesPerSentence = (float)ruleMatchCount / sentenceCount;
        System.out.printf(Locale.ENGLISH, lang + ": ø%.2f rule matches per sentence\n", matchesPerSentence);
        System.out.printf(Locale.ENGLISH, lang + ": %d input lines ignored (e.g. not between %d and %d chars or at least %d tokens)\n", ignoredCount, 
          SentenceSource.MIN_SENTENCE_LENGTH, SentenceSource.MAX_SENTENCE_LENGTH, SentenceSource.MIN_SENTENCE_TOKEN_COUNT);
        if (options.hasOption("print-duration")) {
          System.out.println("The analysis took " + (System.currentTimeMillis() - startTime) + "ms");
        }

View on GitHub (pinned to 2e990059ce)

Solutions

1. Read the truncated sentence text in the message to find the offending content
2. Re-run with --skip-exceptions to print stack traces and continue past bad sentences
3. Fix or disable the specific rule that fails on that sentence
4. Catch and skip in a wrapper around your own check pipeline

Example fix

// before
java ... -l en dump.xml
// after
java ... -l en --skip-exceptions dump.xml

// pre-screen sentences before the checker
if (sentence.getText() == null || sentence.getText().length() > 5000) {
  return; // skip pathological sentences
}

try {
  checker.run(dumpFiles);
} catch (RuntimeException e) {
  if (e.getMessage() != null && e.getMessage().startsWith("Check failed on sentence")) {
    log.warn("Skipping run at sentence: " + e.getMessage(), e);
  } else throw e;
}

When it happens

Trigger: LanguageTool rule code throws on pathological sentence text (regex stack overflow, null handling); the sentence source yields corrupt text; a plugin/rule bug triggered by specific content.

Common situations: StackOverflowError in regex-based rules on very long sentences; NPEs in custom rules; unusual Unicode in Wikipedia articles.

AI-assisted analysis of languagetool-org/languagetool@2e990059ce (2026-09-06). Data as JSON: /api/errors/25ec324082a90d0a. Report an issue: GitHub.