dbt Core
dbt Fusion's second beta is adapter work: ClickHouse gets materializations, indexes, and catalogs
Comparison · Analytics
nanoparquet vs tulpa
A side-by-side editorial comparison of nanoparquet and tulpa — release velocity, themes, recent moves, and the top alternatives to consider.
nanoparquet vs tulpa: at a glance
What is nanoparquet?
nanoparquet is chasing byte-level agreement with the Java and Rust Parquet readers, not feature count.
nanoparquet reads and writes Parquet from R with no Arrow dependency, which is its entire reason to exist. The 0.4.0 line renamed the reader API and added schema authoring plus `append_parquet()`, and the 0.5.x releases have gone after interoperability: definition and repetition level encodings the Apache Parquet Java library expects, flatbuffer alignment the Rust arrow-rs reader expects, 128-bit decimals, and Polars-written files that omit the dictionary page offset. The newest release adds `bit64::integer64` columns and writing to stdout.
What is tulpa?
The 0.0.x train stops at CRAN: tulpa's engine ships to the ecosystem it already anchors.
tulpa is the C++/R Bayesian spatial inference engine sitting under gcol33's family of ecological occupancy packages, tagging 0.0.x releases several times a week. 0.1.0 is its first CRAN release, and the notes state outright that the engine surface is unchanged from 0.0.198 — the work is packaging discipline: local T bindings rebound to n_t/n_times, OpenMP teams capped under R CMD check, the pkgdown deploy narrowed, an aspell dictionary added. The window behind it splits between the S3 generics conversion and numerical-correctness work in the nested-Laplace grid.
nanoparquet vs tulpa: editorial side-by-side
nanoparquet
0.0ANALYTICS
nanoparquet is chasing byte-level agreement with the Java and Rust Parquet readers, not feature count.
◆ Current state
nanoparquet reads and writes Parquet from R with no Arrow dependency, which is its entire reason to exist. The 0.4.0 line renamed the reader API and added schema authoring plus `append_parquet()`, and the 0.5.x releases have gone after interoperability: definition and repetition level encodings the Apache Parquet Java library expects, flatbuffer alignment the Rust arrow-rs reader expects, 128-bit decimals, and Polars-written files that omit the dictionary page offset. The newest release adds `bit64::integer64` columns and writing to stdout.
◆ Where it's heading
Almost every entry since 0.4.0 names another engine — Java, arrow-rs, Polars, Arrow schema metadata — which tells you the maintainers are treating cross-reader fidelity as the product rather than R-side ergonomics. The type system is filling in from the edges: DECIMAL beyond 8 bytes, UUID, FLOAT16 and INTERVAL as raw lists, and now 64-bit integers with an explicit read-type option instead of a silent cast to double. Writing to `:stdout:` points at a second audience, shell pipelines rather than interactive R.
◆ Prediction
The remaining unmapped Parquet types the changelog has been parking in raw-vector lists — FLOAT16 and INTERVAL — are the obvious next targets, following the same pattern by which DECIMAL and UUID graduated to real R types.
tulpa
7.5ANALYTICS
The 0.0.x train stops at CRAN: tulpa's engine ships to the ecosystem it already anchors.
◆ Current state
tulpa is the C++/R Bayesian spatial inference engine sitting under gcol33's family of ecological occupancy packages, tagging 0.0.x releases several times a week. 0.1.0 is its first CRAN release, and the notes state outright that the engine surface is unchanged from 0.0.198 — the work is packaging discipline: local T bindings rebound to n_t/n_times, OpenMP teams capped under R CMD check, the pkgdown deploy narrowed, an aspell dictionary added. The window behind it splits between the S3 generics conversion and numerical-correctness work in the nested-Laplace grid.
◆ Where it's heading
Two moves in nine days point at the same destination: the generics conversion made tulpa extensible by downstream packages, and CRAN admission makes it installable by them. The current cadence — several tags a week, some existing only to record a measurement that produced no code change — does not survive CRAN's submission overhead, so the release rhythm has to slow whether or not the project intends it. The correctness work still clusters on the joint nested-Laplace driver, and 0.1.0 extends the same diagnostics habit with .NL_AXIS_SD_REASONS, a closed vocabulary for an outer axis whose grid does not contain its own posterior mode.
◆ Prediction
Expect tulpaObs to follow tulpa onto CRAN, since it is the consumer whose registrations the engine has spent this window unblocking, and expect the version line to move in larger, less frequent steps now that each one carries a submission.
Alternatives to nanoparquet and tulpa
Other Analytics products tracked by Sparkpulse, ranked by recent ship velocity. Each card links to a full editorial trajectory and lets you pivot into a head-to-head comparison with either nanoparquet or tulpa.
Fulcrum
Fulcrum is betting its whole map stack on Esri, with a hard Google Maps cutoff on September 1.
Holistics
Holistics keeps fencing in the AI layer it spent the summer building.
Dovetail
Dovetail spent July opening doors to other tools and August making its own rooms easier to enter.
AgencyAnalytics
AgencyAI got skills three weeks ago; everything since has been making them routine.
NocoDB
Interfaces gets the permissions layer it needed, one release after launching.
See all nanoparquet alternatives → · See all tulpa alternatives →
Recent activity from nanoparquet and tulpa
Latest ship moves from both products, interleaved chronologically. ⚡ = editorial spark.
- 17h agotulpaFirst CRAN release: engine surface unchanged from 0.0.198 ⚡
- 4d agotulpatulpa_re_aghq() exposes the mode/theta cross-Hessian
- 8d agotulpaDense batched joint path could silently drop a grid cell
- 8d agotulpaCalibration and goodness-of-fit entry points become S3 generics ⚡
- 9d agotulpaCUDA backend had two definitions; link order decided if it ran
- 9d agotulpaHyperparameter bounds now flag when they leave the node range
- 4mo agonanoparquet64-bit integer columns and writing Parquet to stdout
- 4mo agonanoparquetFiles now readable by the Java and Rust Parquet libraries
- 1y agonanoparquetReads Polars files that omit the dictionary page offset
- 1y agonanoparquetDate, FLOAT, and mixed-encoding read fixes
- 1y agonanoparquetSchema authoring and append_parquet arrive with a renamed API ⚡
- 1y agonanoparquetFixes a write_parquet crash
Frequently asked questions
What is the difference between nanoparquet and tulpa?
They serve adjacent needs but don't currently overlap on shipped themes. tulpa is currently shipping more aggressively (velocity 7.5 vs 0.0), with 2 editorial sparks in the last 30 days against 0. See the at-a-glance table above for a side-by-side breakdown of velocity, recent sparks, and editorial themes.
Is nanoparquet better than tulpa?
Sparkpulse doesn't pick a winner — we score release velocity, not feature parity. tulpa is currently shipping more aggressively (velocity 7.5 vs 0.0), with 2 editorial sparks in the last 30 days against 0. For your specific use case, the alternatives sections above list other Analytics products to evaluate alongside.
What are the best alternatives to nanoparquet?
Top nanoparquet alternatives in Analytics are ranked by recent ship velocity. Browse the "nanoparquet alternatives" section above for the current picks, or visit /alternatives/nanoparquet for the full list with editorial commentary on each.
What are the best alternatives to tulpa?
Top tulpa alternatives in Analytics are ranked by recent ship velocity. Browse the "tulpa alternatives" section above for the current picks, or visit /alternatives/tulpa for the full list with editorial commentary on each.