Visivo v1.0.82: Metric and Dimension Publishing, End to End

Visivo v1.0.82 completes metric and dimension publishing end-to-end, closing the loop on the semantic layer. It also hardens Parquet writes for null and decimal columns, makes tables fill and scroll their slots instead of overflowing, and captures dashboard thumbnails automatically when a dashboard is viewed. It is the final 1.x release, the last polish before 2.0.

This is a release about finishing things. Each item on the list takes something that was most of the way there and carries it the rest of the way: publishing becomes truly end-to-end, Parquet writes stop choking on edge-case columns, tables behave inside their containers, and dashboards finally get the thumbnails they always should have had. Taken together, v1.0.82 is the clean, complete 1.x baseline that 2.0 launches from.

What shipped in v1.0.82

v1.0.82 bundles four improvements that each remove a rough edge:

• Metric and dimension publishing, end to end. The semantic-layer publishing loop is now complete: define a metric or dimension and it flows all the way through to where it is consumed.
• Robust Parquet writes for null and decimal columns. The Parquet writer now handles nulls and decimals correctly, so real-world data with missing values and precise numerics round-trips cleanly.
• Tables that fill and scroll their slots. Tables now expand to fill the space allotted to them and scroll within it, instead of overflowing their containers.
• Dashboard thumbnails captured on view. Viewing a dashboard now captures a thumbnail of it automatically, so dashboards get accurate previews without any extra step.

None of these is a flashy headline feature. All of them are the kind of finishing work that makes a tool feel solid. That is exactly what a final 1.x release should be.

Metric and dimension publishing, end to end

The most significant item is completing metric and dimension publishing end-to-end. This is the work that closes the semantic layer loop.

The whole promise of a semantic layer is "define once, consume everywhere." You define a metric like revenue, or a dimension like region, a single time, and every dashboard, chart, and downstream consumer reads that one definition. That promise only holds if there is an unbroken path from the definition to the point of consumption. A semantic layer where a metric is defined but cannot actually be published to where it is needed is a half-built bridge.

v1.0.82 finishes the bridge. Metrics and dimensions now publish end-to-end, which means a definition you write does not stop at being declared. It travels all the way through the system to the surfaces that use it. That is what turns the semantic layer from a place to write definitions into a working single source of truth for metrics.

Why this matters: end-to-end publishing is what makes consistency real rather than aspirational. When the path from definition to consumption is complete and reliable, the same metric resolves to the same SQL and the same number everywhere it appears. Two dashboards cannot disagree about revenue when they are both reading the same published definition. That consistency is the entire reason to have a semantic layer, and v1.0.82 is the release where the plumbing behind it is fully connected.

Robust Parquet writes for null and decimal columns

The second improvement hardens how Visivo writes Parquet. Specifically, it makes writes robust for null and decimal columns.

These two column types are where naive Parquet writers most often fall down, and they are everywhere in real data. Nulls are unavoidable: optional fields, outer joins, late-arriving data, and missing measurements all produce them. Decimals are how you represent money and other exact quantities without the rounding errors that floating point introduces. A BI tool that mishandles either is a BI tool that mishandles real datasets.

Parquet is the workhorse columnar format underneath modern analytics, and Visivo uses it to materialize results efficiently. Hardening the writer for nulls and decimals means that data with missing values and precise numerics now round-trips correctly: it writes without error and reads back with its types and values intact. No silent coercion of a decimal to a float, no choking on a column full of nulls.

This is foundational reliability. You should never have to think about whether your analytics tool can faithfully store a currency column or a column with gaps in it. After v1.0.82, you do not.

Tables that fill and scroll their slots

The third fix is a layout improvement: tables now fill and scroll their slots.

Tables are deceptively hard to lay out well. A table with a few rows should sit neatly in its space. A table with thousands of rows or dozens of columns should not blow out of its container and shove the rest of the dashboard around. The right behavior is for a table to expand to fill the slot it is given, and then scroll within that slot when the data exceeds it.

That is what v1.0.82 delivers. Tables now fill the space allotted to them and scroll inside it, so a wide or tall table stays contained. The dashboard layout you designed holds, regardless of how much data lands in any one table. No more a single oversized table breaking the composition of an entire view.

This is a small change with an outsized effect on how finished a dashboard feels. Layout that holds under real data is the difference between a dashboard that looks designed and one that looks like it leaked. It pairs naturally with the kind of pivot and tabular workflows where tables can get large fast.

Dashboard thumbnails captured on view

The fourth improvement is dashboard thumbnails captured on view. When you view a dashboard, Visivo now automatically captures a thumbnail image of it.

This solves a familiar annoyance. Dashboard listings, galleries, and pickers are far easier to navigate with accurate visual previews than with text alone. But generating those previews has traditionally meant a separate rendering step, stale images, or placeholders that do not match the actual dashboard. Capturing the thumbnail at view time sidesteps all of that: the preview is generated from the real, rendered dashboard, so it is accurate by construction, and it happens automatically without any extra action from you.

The payoff is a more navigable, more polished experience anywhere dashboards are listed. You recognize the dashboard you want by sight, and the thumbnail actually reflects what you will see when you open it.

The last 1.x before 2.0

Put the four together and the theme is clear: v1.0.82 is about completion. Publishing is now end-to-end, so the semantic layer is whole. Parquet writes are robust, so real data is safe. Tables behave, so layouts hold. Thumbnails are automatic, so navigation is clean.

That is exactly what you want from a final 1.x release. It is not the place to introduce risk or sprawl. It is the place to close open loops and harden foundations so the major version that follows can start from solid ground. v1.0.82 completes the semantic-layer publishing loop and ties off the loose ends, setting the stage for the 2.0 line, where Insights and the semantic layer become the center of the product.

If you are on a 1.x release, this is the one to land on before stepping up to 2.0. You can update and explore on your own data via get started, or browse the examples gallery to see published metrics and well-behaved tables in action.

Previously in Visivo

In the previous installment, Visivo v1.0.81: The Redesigned Explorer (Gamma), we promoted the modern Explorer to the default and removed the legacy frontend routes. v1.0.82 continues that pre-2.0 cleanup, finishing the semantic-layer publishing loop and hardening the foundations right before the major version lands.