Inverse logarithmic scale [0,0.9,0.99,0.999]

I'm trying to plot the equivalent of a hdrhistogram, to analyse some latency data, however it seems non-trivial to do so since it requires what is essentially the inverse of a logarithmic scale.

I.e what I am trying to get is a scale that has ticks along the lines of: [0,0.9, 0.99, 0.999, 0.9999]

I'm coding this all via the Altair library for python, if that helps in any way.

Solution

There is no easy way to do this in Altair, because it's not supported in Vega (see the two-year-old feature request here: https://github.com/vega/vega/issues/1277)

But you can hack around it by transforming your data, using a standard log scale, and then computing new tick labels to reflect your underlying data. It might look something like this:

import altair as alt
import pandas as pd

df = pd.DataFrame({
    'x': range(5),
    'y': [0.0001, 0.9, 0.99, 0.999, 0.9999],
})

alt.Chart(df).transform_calculate(
  z = 1 - alt.datum.y  
).mark_line().encode(
  x='x:Q',
  y=alt.Y(
    'z:Q',
    scale=alt.Scale(type='log', reverse=True)),
    axis=alt.Axis(
      values=[1, 0.1, 0.01, 0.001, 0.0001, 0.00001],
      labelExpr="1 - datum.value"),
)