GrammarViz 3.0 (3.0.4 release, 2026)
July 22, 2026 · View on GitHub
GrammarViz 3.0 source code public repository. This code is released under GPL v.2.0.
For the detailed software description, please visit our demo site.
In a nutshell
GrammarViz 3.0 is a software for time series exploratory analysis with GUI and CLI interfaces. The GUI enables interactive time series exploration workflow that allows for variable length recurrent and anomalous patterns discovery from time series [4]:
It is implemented in Java and is based on continuous signal discretization with SAX, Grammatical Inference with Sequitur and Re-Pair, and algorithmic (Kolmogorov) complexity.
Cross-language checks for the shared SAX and GI layers (discord search, sliding-window SAX, RePair) live in jmotif-conformance.
In contrast with 2.0, GrammarViz 3.0 introduces an approach for the grammar rule pruning and the automated discretization parameters selection procedure based on the greedy grammar rule pruning and MDL -- by sampling a possible parameters space, it finds a parameters set which produces the most concise grammar describing the observed time series the best, which often is close to the optimal -- here concise and describing are based on other specific criteria.
What's new in the 3.0.4 release
Memory and allocation lean-down for long series and deep grammars:
- RRA search reuses thread-local distance buffers and a precomputed rule-occurrence index.
- GUI clears chart/log state on reload to avoid retaining large series in memory.
What's new in the 3.0.2 release
GUI modernization and concurrency hardening since 3.0.1:
- Dependencies: dropped abandoned SwingX; upgraded JFreeChart 1.0.19 → 1.5.6.
- MVC: replaced deprecated
Observable/Observerwith a typed listener + thread-safe message board; UI updates centralized on the EDT. - Anomaly search runs on a
SwingWorker— the GUI stays responsive while RRA runs. - Workflow concurrency: load, discretize, prune, and cluster run off the EDT with a single-flight latch; the parameter guesser is single-flight and terminable.
- Fixes: chart-aware anomaly guards, histogram empty-grammar guard, rule-density write off the EDT, save-chart after histogram, PMD + SpotBugs quality gate.
See CHANGELOG.md for the full list.
What's new in the 3.0.1 release
Maintenance release: bumps the jMotif stack to 2.0.1 (jmotif-sax, jmotif-gi), aligns
SLF4J/Logback/JaCoCo with the rest of the family, and installs SAX/GI from source in CI
until jmotif-sax 2.0.1 is on Maven Central. No application behavior changes.
What's new in the 3.0.0 release
This release modernizes the build and fixes a cluster of long-standing defects in the automated parameter-selection ("Guess parameters") workflow. Highlights:
- Builds on Java 21 against the aligned jMotif 2.0.0 libraries
(
jmotif-saxandjmotif-gi2.0.0); CI runs on Java 21 and 25. - The parameter guesser now honors your input. Previously the window/PAA/alphabet ranges, step sizes, and minimal-cover threshold typed into the guesser dialog were silently discarded, and the inclusive maximum of each range was never actually evaluated -- both are fixed, with input validation on the dialog.
- Coverage-aware selection. The guesser now picks the most concise parameter set whose pruned grammar meets your minimal rule-cover threshold, instead of ignoring coverage; if none qualify it falls back gracefully and tells you so.
- No more silent hangs. A degenerate range or sampling error used to leave the UI stuck on "Stop!" indefinitely; it now reports the problem and resets.
- RePair is the default grammar-induction algorithm (selectable in Options, alongside Sequitur).
- First unit-test coverage for the grammar reductor and the parameter selector (26 tests).
See CHANGELOG.md for the full list, including behavior changes for existing users.
It also implements the "Rule Density Curve" and "Rare Rule Anomaly (RRA)" algorithms for time series anomaly discovery [5], that significantly outperform HOT-SAX algorithm for time series discord discovery, which is current state of the art. In the table below, the algorithms performance is measured in the amount of calls to the distance function (less is better). The last column shows the RRA performance improvement over HOT-SAX:
| Dataset and SAX parameters | Dataset size | Brute Force | HOT-SAX | RRA | Reduction |
|---|---|---|---|---|---|
| Daily commute (350,15,4) | 17,175 | 271,442,101 | 879,067 | 112,405 | 87.2% |
| Dutch power demand (750,6,3) | 35,040 | 1.13 * | 6,196,356 | 327,950 | 95.7% |
| ECG 0606 (120,4,4) | 2,300 | 4,241,541 | 72,390 | 16,717 | 76.9% |
| ECG 308 (300,4,4) | 5,400 | 23,044,801 | 327,454 | 14,655 | 95.5% |
| ECG 15 (300,4,4) | 15,000 | 207,374,401 | 1,434,665 | 111,348 | 92.2% |
| ECG 108 (300,4,4) | 21,600 | 441,021,001 | 6,041,145 | 150,184 | 97.5% |
| ECG 300 (300,4,4) | 536,976 | 288 * | 101,427,254 | 17,712,845 | 82.6% |
| ECG 318 (300,4,4) | 586,086 | 343 * | 45,513,790 | 10,000,632 | 78.0% |
| Respiration, NPRS 43 (128,5,4) | 4,000 | 14,021,281 | 89,570 | 45,352 | 49.3% |
| Respiration, NPRS 44 (128,5,4) | 24,125 | 569,753,031 | 1,146,145 | 257,529 | 77.5% |
| Video dataset (150,5,3) | 11,251 | 119,935,353 | 758,456 | 69,910 | 90.8% |
| Shuttle telemetry, TEK14 (128,4,4) | 5,000 | 22,510,281 | 691,194 | 48,226 | 93.0% |
| Shuttle telemetry, TEK16 (128,4,4) | 5,000 | 22,491,306 | 61,682 | 15,573 | 74.8% |
| Shuttle telemetry, TEK17 (128,4,4) | 5,000 | 22,491,306 | 164,225 | 78,211 | 52.4% |
References
[1] Lin, J., Keogh, E., Wei, L. and Lonardi, S., Experiencing SAX: a Novel Symbolic Representation of Time Series. DMKD Journal, 2007.
[2] Nevill-Manning, C.G., Witten, I.H., Identifying Hierarchical Structure in Sequences: A linear-time algorithm. arXiv:cs/9709102, 1997.
[3] Larsson, N. J., Moffat, A., Offline Dictionary-Based Compression, IEEE 88 (11): 1722–1732, doi:10.1109/5.892708, 2000.
Citing this work
[4] Pavel Senin, Jessica Lin, Xing Wang, Tim Oates, Sunil Gandhi, Arnold P. Boedihardjo, Crystal Chen, and Susan Frankenstein. 2018. GrammarViz 3.0: Interactive Discovery of Variable-Length Time Series Patterns. ACM Trans. Knowl. Discov. Data 12, 1, Article 10 (February 2018), 28 pages. DOI: https://doi.org/10.1145/3051126
[5] Senin, P., Lin, J., Wang, X., Oates, T., Gandhi, S., Boedihardjo, A.P., Chen, C., Frankenstein, S., Lerner, M., Time series anomaly discovery with grammar-based compression, The International Conference on Extending Database Technology, EDBT 15.
Building
We use Maven and Java 21 to build an executable. Version 3.0.4 depends on
jmotif-sax 2.0.1 and jmotif-gi 2.0.1, which are not yet on Maven Central —
install them from sibling checkouts first:
$ git clone https://github.com/jMotif/SAX.git ../SAX $ git clone https://github.com/jMotif/GI.git ../GI $ mvn -f ../SAX/pom.xml install -P single -DskipTests $ mvn -f ../GI/pom.xml install -DskipTests $ mvn -Psingle package
The GitHub Actions CI clones and installs SAX/GI before each build. Below is a full build trace on Java 21 after the local installs above:
$ java -version openjdk version "21.0.11" 2026-04-21 OpenJDK Runtime Environment (build 21.0.11+10-1-24.04.2-Ubuntu) OpenJDK 64-Bit Server VM (build 21.0.11+10-1-24.04.2-Ubuntu, mixed mode, sharing) $ mvn -version Apache Maven 3.8.7 Java version: 21.0.11, vendor: Ubuntu, runtime: /usr/lib/jvm/java-21-openjdk-amd64 Default locale: en_US, platform encoding: UTF-8 $ mvn package -Psingle [INFO] Scanning for projects... .... [INFO] ------------------------------------------------------- [INFO] T E S T S [INFO] ------------------------------------------------------- [INFO] Running net.seninp.grammarviz.view.TestGrammarReductor [INFO] Running net.seninp.grammarviz.view.TestParamsSamplerSelection [INFO] Running net.seninp.grammarviz.view.TestParamsSamplerGrid [INFO] Running net.seninp.grammarviz.anomaly.TestRRAanomaly [INFO] Running net.seninp.tinker.TestInterval [INFO] [INFO] Results: [INFO] [INFO] Tests run: 26, Failures: 0, Errors: 0, Skipped: 0 [INFO] [INFO] --- jacoco-maven-plugin:0.8.15:report (report) @ grammarviz2 --- [INFO] Analyzed bundle 'GrammarViz2' with 25 classes [INFO] [INFO] --- maven-assembly-plugin:3.3.0:single (make-assembly) @ grammarviz2 --- [INFO] Building jar: target/grammarviz2-3.0.4.jar [INFO] Building jar: target/grammarviz2-3.0.4-jar-with-dependencies.jar [INFO] ------------------------------------------------------------------------ [INFO] BUILD SUCCESS [INFO] ------------------------------------------------------------------------ [INFO] Total time: 11.205 s [INFO] ------------------------------------------------------------------------
Running
To run the GrammarViz 3.0 GUI use the net.seninp.grammarviz.GrammarVizGUI class, or run the self-contained jar from the command line: $ java -Xmx4g -jar target/grammarviz2-3.0.4-jar-with-dependencies.jar (here I have allocated a max of 4 GB of memory for GrammarViz).
CLI interface
By using CLI as discussed in these tutorials, it is possible to save the inferred grammar, motifs, and discords.
Made with Aloha!
