Changes in version 0.1.0.9000                      

  - whence_trace(), whence_table() and the render functions accept lazy
    = "count" to count each step of a lazy pipeline. Each lazy step
    needs a count query, with an extra query for each inner join. Use
    options(whence.lazy = "count") to opt in by default. Row attribution
    starts from the first materialising step that returns a data frame.

  - whence_trace() follows collect(), compute(), as_tibble() and
    as.data.frame() as hidden steps that keep rows. For lazy sources,
    the last materialising step becomes the source and later steps are
    traced.

  - whence_trace() gives a clear error for lazy tables from 'dbplyr',
    'arrow', 'dtplyr' and prudent 'duckplyr' that still need
    materialisation.

  - whence_trace() traces inner_join(), left_join(), semi_join() and
    anti_join(). Counts show removals and repeated matches separately.
    whence_rows() finds contributing and removed rows of the main
    source.

                 Changes in version 0.1.0 (2026-10-24)                  

  - Initial CRAN submission.

  - whenceOutput() and renderWhence() show a value in 'shiny'. A click
    on the value opens a panel with the attrition waterfall and the
    reproduce code, with no trip to the server. format in renderWhence()
    turns the value into text, and a formatter of 'scales' works there.

  - whenceTableOutput() and renderWhenceTable() show a summary table. A
    click on a row opens the panel of that row: the rows of its group at
    each step, the code of the row, and its rows.

  - whencePlotOutput() and renderWhencePlot() draw a 'ggplot2' plot of a
    summary. A click on a bar or a point opens the panel of its row,
    also when the axis is computed, as in aes(reorder(region, total),
    total).

  - whence_code() gives code that makes the value again. Only input$name
    and input[["name"]] are replaced by their values.

  - whence_rows() finds the source rows that make the value. With step,
    it finds the rows that one step removed. It runs only when it is
    called. In the panel, each count of removed rows, with its
    percentage, is a button that shows them. print() and the panel show
    the number of each step.

  - whence_table() and whence_group() trace a summary table and explain
    one of its rows, with no 'shiny'. whence_group() finds the row by
    its number, or by its group values, such as whence_group(tab, region
    = "APAC").

  - whence_opened() gives a reactive with the panel that the viewer has
    open, and the row of the table, so an app can follow the viewer.

  - whence_trace() splits a pipeline into its steps and records the rows
    after each step. When the chain stops at a call that whence does not
    know, such as a join, the trace and the panel say where.
    pivot_longer() and pivot_wider() from 'tidyr' are known steps.