Slicing

Novem selectors choose rows and columns for styling and formatting. The same notation is used for table cells, chart colours, mark types, labels, facets, and legends.

A slicer can be an index, an inclusive start:stop:step range, or a comma-separated list of indices. Indices start at zero. Negative indices count from the end in ranges and single-index selectors, but are not supported inside comma-separated lists.

Syntax

A selector combines a row slicer and a column slicer, separated by a space.

Any styles using selectors apply to the intersection of the two slices — the block of cells where the selected rows and columns cross.

Slicers

The slicing operation consists of one or more integers separated by a colon :. The first number denotes the starting position (counting from zero), the second the ending position and the third the step (how many items to skip). The integer values can be skipped to indicate start or end of a range.

Selectors

For a complete selector, a slice is specified twice, once for rows and once for columns. The instructions are separated with a space.

Below are a few examples of selectors along with some explanations.

-- select all rows and all columns
  :     :

-- select first row and all columns
 0      :

-- select last row and all columns
-1      :

-- select first three rows
 0:2    :   -- 2 is the third element (zero-indexed), and the stop is INCLUSIVE

-- select last three rows
-3:     :

-- select a 3x3 matrix in the top left corner of the table
 0:2   0:2
  :2    :2  -- the zero is implied if omitted

-- select every other row in the table
  ::2   :

-- select the first and last row
0,-1    :

-- select row 4,5,6
3,4,5   :   -- remember zero index

One index space: the raw file

Every selector addresses the raw uploaded file: (0,0) is the very first value of your CSV. Row 0 is the header line, column 0 the leftmost column — on a table you can style them like any other cell.

Charts use the exact same coordinates. Column 0 is the x /category column and row 0 the header line; neither carries a paintable mark, so a selector that lands there simply paints nothing. In practice: the first series is column 1, the first data point row 1, and a line that works on a table means the same cells on a chart of the same CSV. Pie and donut slices are rows — 2 : addresses the second slice.

Two spaces are deliberately not the CSV, because they only exist after layout: facet strip-label cells and per-panel margins address the physical panel grid, and the axis grammar's third slot addresses positions in the generated tick sequence. They still use the same slicer vocabulary.

One slicer

A line with a single slicer addresses columns: revenue b bolds the revenue column, 1 red colors the first series. Two slicers are always rows × columns, and the two-slicer reading wins whenever both leading tokens parse as slicers.

Header names

A column slicer can use the column's name instead of its index: revenue, big_cat (spaces read as underscores, case-insensitive). Names work inside ranges and lists — revenue:costs, q1,q3 — and resolve to the column's raw position. Numeric tokens are always indices: a header named 2024 does not change the positional meaning of 2024. Row selectors are always positional.

Edge semantics

• Stop is inclusive: 0:2 is three rows (the pandas .loc rule).
• Overflows are inclusive: :1000 on 100 rows is all 100, -500: on 50 rows is all 50. Only a window entirely outside the axis — 100:200 on 5 rows, or a reversed range like 3:1 — selects nothing.
• Negatives count from the end: -1 is the last row; an over-negative single index (-99 on 3 rows) selects nothing.
• Invalid selectors select nothing. A selector that fails to parse causes its line to be ignored. In comma-separated lists, invalid items are ignored and valid items are retained.

Line markers: + and -

In the chart grammars (type, labels, marks, colors) and the table styling and format files, a + or - as the first character of a line, followed by a space, is a line marker:

• + <selector> [overrides] restates the nearest preceding line over a new selection — l:%v le:t then + 3 puts the same label on another series.
• - <selector> erases those cells from every line above it, and later lines can apply styles to those cells again.

Attached signs always belong to the slicer: -1 red colors the last column, - 1 erases column 1. Table styling and format files take the same markers: a bare - or + as the first token of a line is always a marker there too, and the historical tolerant reading of a bare leading - is retired. On a + line whose tokens are all slicer-shaped the tokens read as the new selector; when a slicer-shaped payload is meant (the align dash), write the row slicer explicitly (+ : 1 -).

Interactive example

Change the selectors below to see which cells they select:

Examples

Below are a couple of examples showing how the slicer selection can be used to apply different styles to a novem table.

Slicers and steps

This example shows a static color pattern reminiscent of a picnic blanket using three shades of blue and setting the background color.

Four selectors use the step value to create the pattern. First color alternating rows starting at different offsets, then add alternating columns. The selectors are applied in order, so later entries take precedence.

-- Create a "picnic blanket" using 4 shades of blue
::2  :     bg blue-100  -- color every other row starting at 0
1::2 :     bg blue-200  -- color every other row starting at 1
1::2 1::2  bg blue-300  -- color every other col starting at row 1 col 1
::2  1::2  bg blue-200  -- color every other row starting at row 0 col 1

Index lists

Here we highlight a hierarchical structure in our table by coloring rows in descending shades of gray. This is a common strategy when wanting to preserve space.

Unlike in the picnic example above, here we explicitly choose which entries to highlight with a comma-separated list of rows.

-- Color rows by "hierarchy"
1           : bg gray-500  -- world
2,5,8       : bg gray-400  -- region 1
3,6,9,12,19 : bg gray-300  -- region 2