Expand description
Value counts: how many rows of the view hold each value of one column, and a summary of the column, from one read of that column.
The count is exact: one streamed pass over the column that keeps a count per
value and no rows (crate::chart::chart_data::Tally). A view too large to count at
once, where the sampler can read part of it (one Parquet or IPC file, read a few
row groups at a time), is sampled first instead and says so; counting every row
is then the user’s call. Where the sampler would stream every row anyway, the
exact count is the same read and is what runs.
Structs§
- Line
- One line of the listing: what it stands for, its rows, and the rows of it and every line above it.
- Plan
- A count to run off the UI thread: the view, the column, and how to read it.
- Summary
- The header strip: what can be said of the column from its counts alone.
- Value
Counts - A column’s values counted: every distinct value with its rows, the summary, and the listing in either order.
Enums§
- Line
Kind - What one line of the listing stands for.
- Number
- A number the summary adds up: whole for integer columns, so a large sum keeps every digit.
- Order
- Which way the values are listed.
- Read
- How a count reads the view.
Constants§
- HISTOGRAM_
BINS - Bins of the histogram view. An integer column spanning fewer values than this takes a bin per value instead.
- LARGE_
SAMPLES - A local view this many times the sample size is sampled first, where a sample reads less of it: 10,000,000 rows at the default sample size.
- MAX_
DISTINCT - Distinct values a count keeps before it stops: past this the column is an identifier, and the counts would grow with the table.
- TOP_N
- Values listed one per line; the rest are summed into one
otherline.