Editable DataTable

The DataTable is editable. Like a spreadsheet, it can be used
as an input for controlling models with a variable number
of inputs.

This chapter includes recipes for:

Reading the contents of the DataTable
Filtering out null values
Uploading data
Determining which cell has changed
Adding or removing columns
Adding or removing rows
Clearable, deletable, renamable, and hideable columns
Export DataTable

Predefined Columns

In this example, we initialize a table with 10 blank rows and
a few predefined columns. To retrieve the data, listen to the
data property.

A few notes:
- If you copy and paste data that is larger than the rows, then the
table will expand to contain the contents.
Try it out by copying and pasting this dataset.
- Unlike other spreadsheet programs, the DataTable has a fixed number of
rows. So, if your model has an arbitrary number of parameters
(rows or columns), we recommend initializing your table with a
large number of empty rows and columns.
- When copying data from the table to an external spreadsheet or
between two tables in different tabs, you can choose to include column headers
by adding include_headers_on_copy_paste=True. However, headers are ignored
when copying between two tables in the same tab.

using Dash
using CSV, DataFrames 

app = dash()

params = [
    "Weight", "Torque", "Width", "Height",
    "Efficiency", "Power", "Displacement"
]

app.layout = html_div([
    dash_datatable(
        id="table-editing-simple",
        columns=(
            vcat([Dict("id" =>  "Model", "name" =>  "Model")],
            [Dict("id" =>  p, "name" =>  p) for p in params])
        ),
        data=[Dict("Model" => i, [param => 0
          for param in params]...) for i in 1:5
        ],
        editable=true
    ),
    dcc_graph(id="table-editing-simple-output")
])

callback!(app,
    Output("table-editing-simple-output", "figure"),
    Input("table-editing-simple", "data"),
    Input("table-editing-simple", "columns")
    ) do rows, columns 
    df = DataFrame(rows)
    select!(df, [c.name for c in columns])
    return (
        "data" =>  [Dict(
            "type" =>  "parcoords",
            "dimensions" =>  [Dict(
                "label" =>  col.name,
                "values" =>  df[!, col.id]
            ) for col in columns]
        )]
    )
end

run_server(app, "0.0.0.0", debug=true)

Integration with Dash Loading States

As of table version 4.3.0, Dash loading states also have some
control over whether the table is editable. If the data property
is loading (e.g., while retrieving data from a server), you will
be unable to edit the cells and the dropdowns. This avoids cases
in which the edited value (input from the user) conflicts with the
value that is returned by the server.

In the example below, you can use the dropdown to choose to load
either the style_cell property or the data property. When you
select the property, there will be a simulated delay (to mimic a
delay you might get when communicating with a server). Although
this isn’t the case in the example, you can also use the
.dash-loading CSS selector to apply specific styling while the
table is waiting.

If you select style_cell, you’ll be able to edit the cell as the
value is loading; if you select data, you won’t. Try it out
yourself!

using Dash
using Random

Random.seed!(123)

using CSV, DataFrames 

app = dash()

app.layout = html_div([
    "Choose property to load:  ",
    dcc_dropdown(
        id="loading-states-table-prop",
        options=[
            Dict("label" =>  prop, "value" =>  prop)
            for prop in ["style_cell", "data"]
        ]
    ),
    html_br(),
    dash_datatable(
        id="loading-states-table",
        columns=[Dict(
            "name" =>  "Column $i",
            "id" =>  "column-$i",
            "deletable" =>  true,
            "renamable" =>  true
        ) for i in 1:5],
        data=data=[
          Dict("column-$i" => (rand(1:100)) for i in 1:5)
           for j in 1:5],
        editable=true
    )
])

callback!(app,
    Output("loading-states-table", "style_cell"),
    Input("loading-states-table-prop", "value")
) do value 
    if value == "style_cell"
        sleep(5)
        return Dict("color" =>  "rgb($(rand(0:255)), $(rand(0:255)), $(rand(0:255)))")
    end
    throw(PreventUpdate())
end

callback!(app,
    Output("loading-states-table", "data"),
    Input("loading-states-table-prop", "value")
) do value
    if value == "data"
        sleep(5)
        return [
            Dict("column-$i" => 
             (rand(0:100)) for i in 1:5)
            for j in 1:5
        ]
    end
    throw(PreventUpdate())
end

run_server(app, "0.0.0.0", debug=true)

Choose property to load:

Filtering out Empty Cells

The DataTable will always return all of the cells in the table, even
if the cells haven’t been filled out. So, you’ll likely want to filter
out the empty values.

When you clear a cell, the DataTable will set its contents to ""
(empty string). So, for consistency, we recommend initializing your
empty data with "".

Heads up! In the future, when we introduce proper data types,
we may initialize empty data as something other than "". For example,
if the column is numerical, we’ll want to avoid having any "" in the
data and we may initialize epmty data to `nothing instead.

In this example, we prune away any rows that have empty cells in them.
This is just one way to prune data; you may want to clean your data
differently in your application.

using Dash
using Random

Random.seed!(123)

using CSV, DataFrames 

app = dash()

app.layout = html_div([
dash_datatable(
      id="editing-prune-data",
      columns=[Dict(
          "name" =>  "Column $i",
          "id" =>  "column-$i"
      ) for i in 1:4],
      data=[
          Dict("column-$i" =>  (j + (i-1)*5)-1 for i in 1:4)
          for j in 1:5
      ],
      editable=true
  ),
  html_div(id="editing-prune-data-output")
])

callback!(app,
Output("editing-prune-data-output", "children"),
              Input("editing-prune-data", "data")
) do rows
    pruned_rows = NamedTuple[]
    for row in rows
        # require that all elements in a row are specified
        # the pruning behavior that you need may be different than this
        if all([cell != "" for cell in values(row)])          
            push!(pruned_rows,row)
        end
    end
    return html_div([
        html_div("Raw Data"),
        html_pre(string(rows)),
        html_hr(),
        html_div("Pruned Data"),
        html_pre(string(pruned_rows)),
    ])
end

run_server(app, "0.0.0.0", debug=true)

Uploading Data

A nice recipe is to tie the
<code>dcc.Upload<code>
with the Table component. After the user has uploaded the data, they
could edit the contents or rename the rows.

Here’s an example that creates a simple “x-y” plotter: upload a CSV
with two columns of data and we’ll plot it.
Try it out by downloading this file
and then uploading it.

using Dash
using CSV, DataFrames 
using Base64

external_stylesheets = ["https://codepen.io/chriddyp/pen/bWLwgP.css"]
app = dash(external_stylesheets=external_stylesheets)

app.layout =  html_div([
  dcc_upload(
      id="datatable-upload",
      children=html_div([
          "Drag and Drop or ",
          html_a("Select Files")
      ]),
      style=Dict(
          "width" =>  "100%", "height" =>  "60px", "lineHeight" =>  "60px",
          "borderWidth" =>  "1px", "borderStyle" =>  "dashed",
          "borderRadius" =>  "5px", "textAlign" =>  "center", "margin" =>  "10px"
      ),
  ),
  dash_datatable(id="datatable-upload-container"),
  dcc_graph(id="datatable-upload-graph")
])
function parse_contents(contents, filename)
    content_type, content_string = split(contents, ',')
    decoded = base64decode(content_string)
    df = DataFrame()
    try
      if occursin("csv", filename)
        str = String(decoded)
        df =  CSV.read(IOBuffer(str), DataFrame)
      end
    catch e
      print(e)
      return html_div([
          "There was an error processing this file."
      ])
    end
    return df
  end

callback!(app,
    Output("datatable-upload-container", "data"),
    Output("datatable-upload-container", "columns"),
    Input("datatable-upload", "contents"),
    State("datatable-upload", "filename")
) do contents, filename
    if contents isa Nothing
        return [Dict()], []
    end
    df = parse_contents(contents, filename)
    return Dict.(pairs.(eachrow(df))), [Dict("name" =>  i, "id" =>  i) for i in names(df)]
end

callback!(app,
    Output("datatable-upload-graph", "figure"),
    Input("datatable-upload-container", "data")
) do rows
    if any([isempty(rw) for rw in rows])
        return Dict(
            "data" =>  [Dict(
                "x" =>  [],
                "y" =>  [],
                "type" =>  "bar"
            )]
        )
    end
    df = DataFrame(rows)
    nme = names(df)
    return Dict(
        "data" =>  [Dict(
            "x" =>  df[!, nme[1]],
            "y" =>  df[!, nme[2]],
            "type" =>  "bar"
        )]
    )
end

run_server(app, "0.0.0.0", debug=true)

Adding or Removing Columns

In the DataTable, we’ve provided a built-in UI for deleting columns
but not for adding columns. We recommend using an external button to
add columns.

This is a simple example that plots the data in the spreadsheet as a
heatmap. Try adding or removing columns!

using Dash
using CSV, DataFrames 

app = dash()

app.layout = html_div([
    html_div([
        dcc_input(
            id="editing-columns-name",
            placeholder="Enter a column name...",
            value="",
            style=Dict("padding" =>  10)
        ),
        html_button("Add Column", id="editing-columns-button", n_clicks=0)
    ], style=Dict("height" =>  50)),

    dash_datatable(
        id="editing-columns",
        columns=[Dict(
            "name" =>  "Column $i",
            "id" =>  "column-$i",
            "deletable" =>  true,
            "renamable" =>  true
        ) for i in 1:4],
        data=[
            Dict("column-$i" =>  (j + (i-1)*5)-1 for i in 1:4)
            for j in 1:5
        ],
        editable=true,
    ),
    dcc_graph(id="editing-columns-graph")
])

callback!(app,
    Output("editing-columns", "columns"),
    Input("editing-columns-button", "n_clicks"),
    State("editing-columns-name", "value"),
    State("editing-columns", "columns")
)  do n_clicks, value, existing_columns
    if n_clicks > 0 && value != ""
        push!(existing_columns, Dict(
            "id" =>  value, "name" =>  value,
            "renamable" =>  true, "deletable" =>  true
        ))
    end
    return existing_columns
end

callback!(app,
    Output("editing-columns-graph", "figure"),
    Input("editing-columns", "data"),
    Input("editing-columns", "columns")
) do rows, columns
    try
        return Dict(
            "data" =>  [Dict(
                "type" =>  "heatmap",
                "z" =>  [[row[Symbol(c.id)] for c in columns] for row in rows],
                "x" =>  [c.name for c in columns]
            )]
        )
    catch
        throw(PreventUpdate())
    end
end

run_server(app, "0.0.0.0", debug=true)

Adding or Removing Rows

Similarly as columns, the DataTable has a built-in UI for removing rows
but not for adding rows. You can add rows to the table through an
external button.

using Dash
using CSV, DataFrames 

app = dash()

app.layout = html_div([
    html_div([
        dcc_input(
            id="adding-rows-name",
            placeholder="Enter a column name...",
            value="",
            style=Dict("padding" =>  10)
        ),
        html_button("Add Column", id="adding-rows-button", n_clicks=0)
    ], style=Dict("height" =>  50)),

    dash_datatable(
        id="adding-rows-table",
        columns=[Dict(
            "name" =>  "Column $i",
            "id" =>  "column-$i",
            "deletable" =>  true,
            "renamable" =>  true
        ) for i in 1:4],
        data=[
            Dict("column-$i" =>  (j + (i-1)*5)-1 for i in 1:4)
            for j in 1:5
        ],
        editable=true,
        row_deletable=true
    ),

    html_button("Add Row", id="editing-rows-button", n_clicks=0),

    dcc_graph(id="adding-rows-graph")
])

callback!(app,
    Output("adding-rows-table", "data"),
    Input("editing-rows-button", "n_clicks"),
    State("adding-rows-table", "data"),
    State("adding-rows-table", "columns")
) do n_clicks, rows, columns
    if n_clicks > 0
        push!(rows, Dict(c.id =>  "" for c in columns))
    end
    return rows
end

callback!(app,
    Output("adding-rows-table", "columns"),
    Input("adding-rows-button", "n_clicks"),
    State("adding-rows-name", "value"),
    State("adding-rows-table", "columns")
) do n_clicks, value, existing_columns
    if (n_clicks > 0) && (value != "")
        push!(existing_columns, Dict(
            "id" =>  value, "name" =>  value,
            "renamable" =>  true, "deletable" =>  true
        ))
    end
    return existing_columns
end

callback!(app,
    Output("adding-rows-graph", "figure"),
    Input("adding-rows-table", "data"),
    Input("adding-rows-table", "columns")
) do rows, columns
    try
        return Dict(
            "data" =>  [Dict(
                "type" =>  "heatmap",
                "z" =>  [[row[Symbol(c.id)] for c in columns] for row in rows],
                "x" =>  [c.name for c in columns]
            )]
        )
    catch
        throw(PreventUpdate())
    end
end

run_server(app, "0.0.0.0", debug=true)

Updating Columns of the Same Table

One neat application of DataTable is being able to update the table itself
when you edit cells.

One of the limitations in Dash is that a callback’s Output can’t be
the same as the Input (circular dependencies aren’t supported yet).
So, we couldn’t have Output('table', 'data') and
Input('table', 'data') in the same callback!().

However, we can work around this by using State('table', 'data')
and triggering the callback with Input('table', 'data_timestamp').

This example mimics a traditional spreadsheet like Excel by computing
certain columns based off of other other columns.

using Dash
using CSV, DataFrames 

app = dash()

app.layout = html_div([
    dash_datatable(
        id="computed-table",
        columns=[
            Dict("name" =>  "Input Data", "id" =>  "input-data"),
            Dict("name" =>  "Input Squared", "id" =>  "output-data")
        ],
        data=[Dict("input-data" =>  i) for i in 1:11],
        editable=true,
    ),
])

callback!(app,
    Output("computed-table", "data"),
    Input("computed-table", "data_timestamp"),
    State("computed-table", "data")
    ) do timestamp, rows
    rows_dict =  map(r->Dict(keys(r) .=> values(r)),rows)
    for row in rows_dict
        try
            row[Symbol("output-data")] = Float64(row[Symbol("input-data")])^2
        catch
            row[Symbol("input-data")] = "NA"
        end
    end
    return rows_dict
end

run_server(app, "0.0.0.0", debug=true)

Modifying the Data Table Content

Columns in the table can be hidden, deleted, cleared, and renamed. Each of these actions
are represented by a small icon in the column headers. If there is more than one header row,
you can choose where the icons appear. If you want to override these icons, you can do so
by using table css selectors, as demonstrated by the example.

When the clear or delete action is performed, the associated filters are also cleared.
Hiding or deleting can only be done if there are more than one column left in the
table after the action is performed.

In this example, we have included a variety of column actions. Try:

Clear the first column: the content is cleared (or multiple columns when headers are merged)
without deleting the column itself.
Delete the second column: column is deleted from the table and can’t be seen again unless the
page is refreshed.
Rename the third column: the content of selected column headers is edited.
Hide the fourth column: the entire column is hidden from view and can be made visible again
through the toggle columns button.

using Dash
using CSV, DataFrames 

app = dash()

app.layout = dash_datatable(
    columns=[
        Dict("name" =>  ["", "Year"], "id" =>  "year", "clearable" =>  "first" ),
        Dict("name" =>  ["City", "Montreal"], "id" =>  "montreal", "deletable" =>  [false, true]),
        Dict("name" =>  ["City", "Toronto"], "id" =>  "toronto", "renamable" =>  true ),
        Dict("name" =>  ["City", "Ottawa"], "id" =>  "ottawa", "hideable" =>  "last"),
        Dict("name" =>  ["City", "Vancouver"], "id" =>  "vancouver", "clearable" =>  true, "renamable" =>  true, "hideable" =>  true, "deletable" =>  true ),
        Dict("name" =>  ["Climate", "Temperature"], "id" =>  "temp"),
        Dict("name" =>  ["Climate", "Humidity"], "id" =>  "humidity"),
    ],
    data=[
        Dict(
            "year" =>  i,
            "montreal" =>  i * 10,
            "toronto" =>  i * 100,
            "ottawa" =>  i * -1,
            "vancouver" =>  i * -10,
            "temp" =>  i * -100,
            "humidity" =>  i * 5,
        )
        for i in 1:10
    ],
    css=[
        Dict("selector" =>  ".column-header--delete svg", "rule" =>  "display:  none"),
        Dict("selector" =>  ".column-header--delete::before", "rule" =>  "content: X")
    ]
)

run_server(app, "0.0.0.0", debug=true)

Exporting Data Table

The table data can be exported either as csv or xlsx file. You can customize table
headers in the export file. Headers can be column IDs, names, or as displayed.
The difference between export_headers: name and export_headers: display is that you have
the option to download the table with merged headers if your table headers are merged.
Finally, if an action was performed on columns (cleared, deleted,hidden, sorted, filtered), then
the downloaded table will display the current view.

Note that display mode is only supported for export_format: xlsx due to the fact that
headers in csv files can not be merged.

using Dash
using CSV, DataFrames 

app = dash()

app.layout = dash_datatable(
    columns=[
        Dict("name" =>  ["", "Year"], "id" =>  "year" ),
        Dict("name" =>  ["City", "Montreal"], "id" =>  "montreal", "deletable" =>  [false, true]),
        Dict("name" =>  ["City", "Toronto"], "id" =>  "toronto", "renamable" =>  true ),
        Dict("name" =>  ["City", "Ottawa"], "id" =>  "ottawa", "hideable" =>  "last"),
        Dict("name" =>  ["City", "Vancouver"], "id" =>  "vancouver"),
        Dict("name" =>  ["Climate", "Temperature"], "id" =>  "temp"),
        Dict("name" =>  ["Climate", "Humidity"], "id" =>  "humidity"),
    ],
    data=[
        Dict(
            "year" =>  i,
            "montreal" =>  i * 10,
            "toronto" =>  i * 100,
            "ottawa" =>  i * -1,
            "vancouver" =>  i * -10,
            "temp" =>  i * -100,
            "humidity" =>  i * 5,
        )
        for i in 1:10
    ],
    export_format="xlsx",
    export_headers="display",
    merge_duplicate_headers=true
)

run_server(app, "0.0.0.0", debug=true)