Taxinomia is a table oriented analytics system.

Next features

• sorting
• aggregation
• display limit for grouped tables
• grouping function (can be done with computed columns)
• filtering function (can be done with computed columns)
• column optimisations
• sub columns, e.g. a column consists of a proto message, expand it on demand
• when by definition the value of two columns always match, display the value of the other columns when the first one is grouped
• support extremely large tables - user will have to filter on specific columns to reduce the table before being able to load it, though the column might be grouped
• break up a column in multiple columns in one go, date => year/quarter/month/week/day/hour, basically a split like function that generates multiple computed columns automatically
• graphs, etc

CLAUDE.md

Scalability goals are to support up to 1'000'000'000 rows Latency must be near impercetible for up to 1'000'000 rows, maybe 100 ms Latency must be small for up to 10'000'000 rows, maybe 1 s

The UI must be simple and clean, no fancy features, must be intuitive. Avoid following heavy handed approaches that are far too common. Think hard before proposing drop-downs, ask if no obvious alternative is available.

Minimize client-side code, everything is encoded in the URL, updates must always go through the back end, with some very limited exceptions only, These should be clearly documented, justsified and validated.

There is no point in displaying more than at the very most a few hundreds of rows, typically it should not be more than 20 - 100.

Everything in the displayed data must be fully deterministic, page refreshes should change the content of the page

The code should be as simple as possible and cleanly implemented with proper abstractions. Avoid overcomplicating the code, in case of doubts ask.

The View Model must not perform any data computation. Ask if there is a case.

Column types

• base table
• joined columns
• derived columns

Aggregation

• numbers: sum, average, std,
• strings: first, last
• bool: all, none, some/any, count

Filtering

• Any column can be filtered on and/or grouped

Sorting

• on columns in any order between columns
• on grouped values and aggregates

Some implementation gaps

• further joins
• filtering on joined columns
• grouping of joined columns

Future features

• filtering on aggregated values
• grouping on aggregated values ???
• pivoting
• materialization

Expression

• Filtering expression
• Grouping expression
• Expression for derived columns Same syntax and same interpreter

Importing data

• protobuf
• csv
• json, ndjson
• xml
• parquet
• Big Query

Data types

• integers (uint8, 32 64, int8 32 64)
• floats (single and double)
• date and time and duration
• string
• enums (?)
• bool

Active North A 1 1 Active North A 2 3 Active North B 3 5 Active South A 4 3 Active South B 5 2 Active South B 6 2 Active East Z 6 2 Active East C 6 2 Active East C 6 2 Inactive North A 1 1 Inactive North A 2 3 Inactive North B 3 5 Inactive South A 4 3 Inactive South B 5 2 Inactive South B 6 2 Inactive East Z 6 2 Inactive East A 6 2 Inactive East Z 6 2 Pending North A 1 1 Pending North A 2 3 Pending North A 3 5 Pending South A 4 3 Pending South B 5 2 Pending South B 6 2 Pending East Z 6 2 Pending East C 6 2 Pending East C 6 2

|-----------|---------|---| | | North | A | | | |---| | | | B | | |---------|---| | Active | South | A | | | |---| | | | B | | |---------|---| | | East | C | | | |---|