* TODO Explicitly typed columns
Having FatTable infer column types from the content of a table is nice, but as
I learned from the need to have force_string!, sometimes we want to restrict
the type of the column and throw an error if the content does not conform.
So, perhaps when the table columns are specified, we can allow them to be
terminated with a "sigil" that restricts its type:

- @ :: at Date or DateTime column,
- # :: a Numeric column
- ? :: a Boolean column
- ~ :: a String Column,
- * :: a tolerant Column, i.e., it infers the type from the content, but
  allows strings to be added as items in an otherwise typed column but treat
  them a nil for calculation purposes.
- @*, #*, ?* :: a Date, Numeric, or Boolean column that is also tolerant.
  `~*` could be treated the same as `~`.

* TODO Specify Column Widths
Allow a formatter to specify column widths.  This could be a number of
characters, which would be interpreted as a number of "ems" for LaTeX.
Cell content larger than the width would be truncated.  Any column without a
width specified would be set at the width of the longest value in that cell,
after initial formatting.

#+begin_SRC ruby
  tab.to_text do |f|
    f.widths(a: 13, b: 30)
  end
#+end_SRC

Possible enhancements:
- specify an overall width and column widths as decimal or fractions, so that
  a column's width would be that fraction of the overall width.
- specify a Range for a width, so that the column would at least min and at
  most max, otherwise the width of its largest cell.

* TODO Conversion to Spreadsheets
- State "TODO"       from              [2017-04-21 Fri 10:36]
This is a [[https://github.com/westonganger/spreadsheet_architect][gem]] that I can include into the Table model to convert a table into
a spread-sheet, or even a sheet in a multi-sheet spreadsheet file.

* TODO Conversion to HTML
Perhaps an obvious formatter would be HTML tables, but CSS controls the
styling, so have to think on that one.

* TODO Conversion to JSON, YAML
These would need no color formatting and might prove useful.

* TODO Add Quandl or EODDATA Queries
Possible replacements for YQL.

* CNCL Add from_yql for fetching from Yahoo
CLOSED: [2022-01-30 Sun 06:03]
- State "TODO"       from              [2017-04-21 Fri 10:35]

Cancelled because Yahoo shut down the YQL api service.

Add a constructor to allow fetching stock data from yql.  Perhaps grab all
available fields, then allow a select of those of interest.

* DONE Allow sorting by expression
CLOSED: [2022-01-20 Thu 12:47]
Either by a single string argument as the sole argument to order_by, or use
another method, such a order_with.  Note that this can be done now by creating
a new column having the sort expression with select and then just order_by
that column.  Perhaps that is an easy way to implement it.

* DONE Ensure that columns resulting from aggregates have proper type
CLOSED: [2017-12-29 Fri 05:34]
- State "WAIT"       from "TODO"       [2017-12-29 Fri 05:34]
- State "TODO"       from              [2017-11-27 Mon 04:46]
After applying avg, does the column have the proper Numeric or Date, or DateTime
type.  How about Boolean aggregates?

* DONE Formatters
CLOSED: [2017-04-21 Fri 10:36]
- State "WAIT"       from "TODO"       [2017-04-21 Fri 10:36]
- State "TODO"       from              [2017-04-21 Fri 10:35]
Need to think about ways to define formatters for Table for different output
types, including tty, color-tty, latex, csv, spreadsheet?

* DONE Add a Group Boundary concept
CLOSED: [2017-04-21 Fri 10:36]
- State "WAIT"       from "TODO"       [2017-04-21 Fri 10:36]
- State "TODO"       from              [2017-04-21 Fri 10:35]
If I want a table to perform sub-totals at various break points, need to have a
way for a table to record its grouping boundaries. Maybe an array of row
numbers?  Automatically injected by the group-by method?

* DONE Add uniq method and set operations
CLOSED: [2017-03-02 Thu 15:54]
- State "WAIT"       from "TODO"       [2017-03-02 Thu 15:54]
- State "TODO"       from              [2017-03-02 Thu 15:54]
For tables, add a method that eliminates any duplicate rows. Perhaps just apply
Array#uniq to the columns?