mini-align.txt

==============================================================================
------------------------------------------------------------------------------
                                                                    *mini.align*
                                                                     *MiniAlign*
Align text interactively (with or without instant preview). Allows rich and
flexible customization of both alignment rules and user interaction. Works
with charwise, linewise, and blockwise selections in both Normal mode (on
textobject/motion; with dot-repeat) and Visual mode.

Features:
- Alignment is done in three main steps:
    - <Split> lines into parts based on Lua pattern(s) or user-supplied rule.
    - <Justify> parts for certain side(s) to be same width inside columns.
    - <Merge> parts to be lines, with customizable delimiter(s).
  Each main step can be preceded by other steps (pre-steps) to achieve
  highly customizable outcome. See `steps` value in |MiniAlign.config|. For
  more details, see |MiniAlign-glossary| and |MiniAlign-algorithm|.
- User can control alignment interactively by pressing customizable modifiers
  (single keys representing how alignment steps and/or options should change).
  Some of default modifiers:
    - Press `s` to enter split Lua pattern.
    - Press `j` to choose justification side from available ones ("left",
      "center", "right", "none").
    - Press `m` to enter merge delimiter.
    - Press `f` to enter filter Lua expression to configure which parts
      will be affected (like "align only first column").
    - Press `i` to ignore some commonly unwanted split matches.
    - Press `p` to pair neighboring parts so they be aligned together.
    - Press `t` to trim whitespace from parts.
    - Press `<BS>` (backspace) to delete some last pre-step.
  For more details, see |MiniAlign-modifiers-builtin| and |MiniAlign-examples|.
- Alignment can be done with instant preview (result is updated after each
  modifier) or without it (result is shown and accepted after non-default
  split pattern is set).
- Every user interaction is accompanied with helper status message showing
  relevant information about current alignment process.

# Setup~

This module needs a setup with `require('mini.align').setup({})` (replace
`{}` with your `config` table). It will create global Lua table `MiniAlign`
which you can use for scripting or manually (with `:lua MiniAlign.*`).

See |MiniAlign.config| for available config settings.

You can override runtime config settings (like `config.modifiers`) locally
to buffer inside `vim.b.minialign_config` which should have same structure
as `MiniAlign.config`. See |mini.nvim-buffer-local-config| for more details.

# Comparisons~

- 'junegunn/vim-easy-align':
    - 'mini.align' is mostly designed after 'junegunn/vim-easy-align', so
      there are a lot of similarities.
    - Both plugins allow users to change alignment options interactively by
      pressing modifier keys (albeit completely different default ones).
      'junegunn/vim-easy-align' has those modifiers fixed, while 'mini.align'
      allows their full customization. See |MiniAlign.config| for examples.
    - 'junegunn/vim-easy-align' is designed to treat delimiters differently
      than other parts of strings. 'mini.align' doesn't distinguish split
      parts from one another by design: splitting is allowed to be done
      based on some other logic than by splitting on delimiters.
    - 'junegunn/vim-easy-align' initially aligns by only first delimiter.
      'mini.align' initially aligns by all delimiter.
    - 'junegunn/vim-easy-align' implements special filtering by delimiter
      row number. 'mini.align' has builtin filtering based on Lua code
      supplied by user in modifier phase. See |MiniAlign.gen_step.filter|
      and 'f' builtin modifier.
    - 'mini.align' treats any non-registered modifier as a plain delimiter
      pattern, while 'junegunn/vim-easy-align' does not.
    - 'mini.align' exports core Lua function used for aligning strings
      (|MiniAlign.align_strings()|).
- 'godlygeek/tabular':
    - 'godlygeek/tabular' is mostly designed around single command which is
      customized by printing its parameters. 'mini.align' implements
      different concept of interactive alignment through pressing
      customizable single character modifiers.
    - 'godlygeek/tabular' can detect region upon which alignment can be
      desirable. 'mini.align' does not by design: use Visual selection or
      textobject/motion to explicitly define region to align.

# Disabling~

To disable, set `g:minialign_disable` (globally) or `b:minialign_disable`
(for a buffer) to `v:true`. Considering high number of different scenarios
and customization intentions, writing exact rules for disabling module's
functionality is left to user. See |mini.nvim-disabling-recipes| for common
recipes.

------------------------------------------------------------------------------
                                                            *MiniAlign-glossary*
Glossary

PARTS   2d array of strings (array of arrays of strings).
        See more in |MiniAlign.as_parts()|.

ROW     First-level array of parts (like `parts[1]`).

COLUMN  Array of strings, constructed from parts elements with the same
        second-level index (like `{ parts[1][1],` `parts[2][1], ... }`).

STEP    A named callable. See |MiniAlign.new_step()|. When used in terms of
        alignment steps, callable takes two arguments: some object (parts
        or string array) and option table.

SPLIT   Process of taking array of strings and converting it into parts.

JUSTIFY Process of taking parts and converting them to aligned parts (all
        elements have same widths inside columns).

MERGE   Process of taking parts and converting it back to array of strings.
        Usually by concatenating rows into strings.

REGION  Table representing region in a buffer. Fields: <from> and <to> for
        inclusive start and end positions (<to> might be `nil` to describe
        empty region). Each position is also a table with line <line> and
        column <col> (both start at 1).

MODE    Either charwise ("char", `v`, |charwise|), linewise ("line", `V`,
        |linewise|) or blockwise ("block", `<C-v>`, |blockwise-visual|)

------------------------------------------------------------------------------
                                                           *MiniAlign-algorithm*
Algorithm design

There are two main processes implemented in 'mini.align': strings alignment
and interactive region alignment. See |MiniAlign-glossary| for more information
about used terms.

Strings alignment ~

Main implementation is in |MiniAlign.align_strings()|. Its input is array of
strings and output - array of aligned strings. The process consists from three
main steps (split, justify, merge) which can be preceded by any number of
preliminary steps (pre-split, pre-justify, pre-merge).

Algorithm:
- <Pre-split>. Take input array of strings and consecutively apply all
  pre-split steps (`steps.pre_split`). Each one has `(strings, opts)` signature
  and should modify array in place.
- <Split>. Take array of strings and convert it to parts with `steps.split()`.
  It has `(strings, opts)` signature and should return parts.
- <Pre-justify>. Take parts and consecutively apply all pre-justify
  steps (`steps.pre_justify`). Each one has `(parts, opts)` signature and
  should modify parts in place.
- <Justify>. Take parts and apply `steps.justify()`. It has `(parts, opts)`
  signature and should modify parts in place.
- <Pre-merge>. Take parts and consecutively apply all pre-merge
  steps (`steps.pre_merge`). Each one has `(parts, opts)` signature and
  should modify parts in place.
- <Merge>. Take parts and convert it to array of strings with `steps.merge()`.
  It has `(parts, opts)` signature and should return array of strings.

Notes:
- All table objects are initially copied so that modification in place doesn't
  affect workflow.
- Default main steps are designed to be controlled via options. See
  |MiniAlign.align_strings()| and default step entries in |MiniAlign.gen_step|.
- All steps are guaranteed to take same option table as second argument.
  This allows steps to "talk" to each other, i.e. earlier steps can pass data
  to later ones.

Interactive region alignment ~

Interactive alignment is a main entry point for most users. It can be done
in two flavors:
- <Without preview>. Initiated via mapping defined in `start` of
  `MiniAlign.config.mappings`. Alignment is accepted once split pattern becomes
  non-default.
- <With preview>. Initiated via mapping defined in `start_with_preview` of
  `MiniAlign.config.mappings`. Alignment result is shown after every modifier
  and is accepted after `<CR>` (`Enter`) is hit. Note: each preview is done by
  applying current alignment steps and options to the initial region lines,
  not the ones currently displaying in preview.

Lifecycle (assuming default mappings):
- <Initiate alignment>:
    - In Normal mode type `ga` (or `gA` to show preview) followed by textobject
      or motion defining region to be aligned.
    - In Visual mode select region and type `ga` (or `gA` to show preview).
  Strings contained in selected region will be used as input to
  |MiniAlign.align_strings()|.
  Beware of mode when selecting region: charwise (`v`), linewise (`V`), or
  blockwise (`<C-v>`). They all behave differently.
- <Press modifiers>. Press single keys one at a time:
    - If pressed key is among table keys of `modifiers` table of
      |MiniAlign.config|, its function value is executed. It usually modifies
      some options(s) and/or affects some pre-step(s).
    - If pressed key is not among defined modifiers, it is treated as plain
      split pattern.
  This process can either end by itself (usually in case of no preview and
  non-default split pattern being set) or you can choose to end it manually.
- <Accept or discard>. In case of active preview, accept current result by
  pressing `<CR>`. Discard any result and return to initial regions with
  either `<Esc>` or `<C-c>`.

See more in |MiniAlign-modifiers-builtin| and |MiniAlign-examples|.

Notes:
- Visual blockwise selection works best with 'virtualedit' equal to "block"
  or "all".

------------------------------------------------------------------------------
                                                   *MiniAlign-modifiers-builtin*
Overview of builtin modifiers

All examples assume interactive alignment with preview in linewise mode. With
default mappings, use `V` to select lines and `gA` to initiate alignment. It
might be helpful to copy lines into modifiable buffer and experiment yourself.

Notes:
- Any pressed key which doesn't have defined modifier will be treated as
  plain split pattern.
- All modifiers can be customized inside |MiniAlign.setup|. See "Modifiers"
  section of |MiniAlign.config|.

Main option modifiers ~

<s> Enter split pattern (confirm prompt by pressing `<CR>`). Input is treated
    as plain delimiter.

    Before: >
    a-b-c
    aa-bb-cc
<
    After typing `s-<CR>`: >
    a -b -c
    aa-bb-cc
<
<j> Choose justify side. Prompts user (with helper message) to type single
    character identifier of side: `l`eft, `c`enter, `r`ight, `n`one.

    Before: >
    a_b_c
    aa_bb_cc
<
    After typing `_jr` (first make split by `_`): >
     a_ b_ c
    aa_bb_cc
<
<m> Enter merge delimiter (confirm prompt by pressing `<CR>`).

    Before: >
    a_b_c
    aa_bb_cc
<
    After typing `_m--<CR>` (first make split by `_`): >
    a --_--b --_--c
    aa--_--bb--_--cc
<
Modifiers adding pre-steps ~

<f> Enter filter expression. See more details in |MiniAlign.gen_step.filter()|.

    Before: >
    a_b_c
    aa_bb_cc
<
    After typing `_fn==1<CR>` (first make split by `_`): >
    a _b_c
    aa_bb_cc
<
<i> Ignore some split matches. It modifies `split_exclude_patterns` option by
    adding commonly wanted patterns. See more details in
    |MiniAlign.gen_step.ignore_split()|.

    Before: >
    /* This_is_assumed_to_be_comment */
    a"_"_b
    aa_bb
<
    After typing `_i` (first make split by `_`): >
    /* This_is_assumed_to_be_comment */
    a"_"_b
    aa  _bb
<
<p> Pair neighboring parts.

    Before: >
    a_b_c
    aaa_bbb_ccc
<
    After typing `_p` (first make split by `_`): >
    a_  b_  c
    aaa_bbb_ccc
<
<t> Trim parts from whitespace on both sides (keeping indentation).

    Before: >
    a   _   b   _   c
      aa _bb _cc
<
    After typing `_t` (first make split by `_`): >
    a   _b _c
      aa_bb_cc
<
Delete some last pre-step ~

<BS> Delete one of the pre-steps. If there is only one kind of pre-steps,
     remove its latest added one. If not, prompt user to choose pre-step kind
     by entering single character: `s`plit, `j`ustify, `m`erge.

     Examples:
     - `tp<BS>` results in only "trim" step to be left.
     - `it<BS>` prompts to choose which step to delete (pre-split or
       pre-justify in this case).

Special configurations for common splits ~

<=> Use special pattern to align by a group of consecutive "=". It can be
    preceded by any number of punctuation marks and followed by some sommon
    punctuation characters. Trim whitespace and merge with single space.

    Before: >
    a=b
    aa<=bb
    aaa===bbb
    aaaa   =   cccc
<
    After typing `=`: >
    a    =   b
    aa   <=  bb
    aaa  === bbb
    aaaa =   cccc
<
<,> Besides splitting by "," character, trim whitespace, pair neighboring
    parts and merge with single space.

    Before: >
    a,b
    aa,bb
    aaa    ,    bbb
<
    After typing `,`: >
    a,   b
    aa,  bb
    aaa, bbb
<
< > (Space bar) Squash consecutive whitespace into single single space (accept
    possible indentation) and split by `%s+` pattern (keeps indentation).

    Before: >
    a b c
      aa    bb   cc
<
    After typing `<Space>`: >
      a  b  c
      aa bb cc

------------------------------------------------------------------------------
                                                            *MiniAlign-examples*
More complex examples to explore functionality

Copy lines in modifiable buffer, initiate alignment with preview (`gAip`)
and try typing suggested key sequences.
These are modified examples taken from 'junegunn/vim-easy-align'.

Equal sign ~

Lines:

# This=is=assumed=to be a comment
"a ="
a =
a = 1
bbbb = 2
ccccccc = 3
ccccccccccccccc
ddd = 4
eeee === eee = eee = eee=f
fff = ggg += gg &&= gg
g != hhhhhhhh == 888
i   := 5
i     %= 5
i       *= 5
j     =~ 5
j   >= 5
aa      =>         123
aa <<= 123
aa        >>= 123
bbb               => 123
c     => 1233123
d   =>      123
dddddd &&= 123
dddddd ||= 123
dddddd /= 123
gg <=> ee

Key sequences:
- `=`
- `=jc`
- `=jr`
- `=m!<CR>`
- `=p`
- `=i` (execute `:lua vim.o.commentstring = '# %s'` for full experience)
- `=<BS>`
- `=<BS>p`
- `=fn==1<CR>`
- `=<BS>fn==1<CR>t`
- `=frow>7<CR>`


------------------------------------------------------------------------------
                                                             *MiniAlign.setup()*
                          `MiniAlign.setup`({config})
Module setup

Parameters~
{config} `(table|nil)` Module config table. See |MiniAlign.config|.

Usage~
`require('mini.align').setup({})` (replace `{}` with your `config` table)

------------------------------------------------------------------------------
                                                              *MiniAlign.config*
                               `MiniAlign.config`
Module config

Default values:
>
  MiniAlign.config = {
    -- Module mappings. Use `''` (empty string) to disable one.
    mappings = {
      start = 'ga',
      start_with_preview = 'gA',
    },

    -- Modifiers changing alignment steps and/or options
    modifiers = {
      -- Main option modifiers
      ['s'] = --<function: enter split pattern>,
      ['j'] = --<function: choose justify side>,
      ['m'] = --<function: enter merge delimiter>,

      -- Modifiers adding pre-steps
      ['f'] = --<function: filter parts by entering Lua expression>,
      ['i'] = --<function: ignore some split matches>,
      ['p'] = --<function: pair parts>,
      ['t'] = --<function: trim parts>,

      -- Delete some last pre-step
      ['<BS>'] = --<function: delete some last pre-step>,

      -- Special configurations for common splits
      ['='] = --<function: enhanced setup for '='>,
      [','] = --<function: enhanced setup for ','>,
      [' '] = --<function: enhanced setup for ' '>,
    },

    -- Default options controlling alignment process
    options = {
      split_pattern = '',
      justify_side = 'left',
      merge_delimiter = '',
    },

    -- Default steps performing alignment (if `nil`, default is used)
    steps = {
      pre_split = {},
      split = nil,
      pre_justify = {},
      justify = nil,
      pre_merge = {},
      merge = nil,
    },
  }
<
# Options ~

## Modifiers ~

`MiniAlign.config.modifiers` is used to define interactive user experience
of managing alignment process. It is a table with single character keys and
modifier function values.

Each modifier function:
- Is called when corresponding modifier key is pressed.
- Has signature `(steps, opts)` and should modify any of its input in place.

Examples:
- Modifier function used for default 'i' modifier:
>
  function(steps, _)
    table.insert(steps.pre_split, MiniAlign.gen_step.ignore_split())
  end
<
- Tweak 't' modifier to use highest indentation instead of keeping it:
>
  require('mini.align').setup({
    t = function(steps, _)
      table.insert(steps.pre_justify, MiniAlign.gen_step.trim('both', 'high'))
    end
  })
<
- Tweak `j` modifier to cycle through available "justify_side" option
  values (like in 'junegunn/vim-easy-align'):
>
  require('mini.align').setup({
    modifiers = {
      j = function(_, opts)
        local next_option = ({
          left = 'center', center = 'right', right = 'none', none = 'left',
        })[opts.justify_side]
        opts.justify_side = next_option or 'left'
      end,
    },
  })
<
## Options ~

`MiniAlign.config.options` defines default values of options used to control
behavior of steps.

Examples:
- Set `justify_side = 'center'` to center align at initialization.

For more details about options see |MiniAlign.align_strings()| and entries of
|MiniAlign.gen_step| for default main steps.

## Steps ~

`MiniAlign.config.steps` defines default steps to be applied during
alignment process.

Examples:
- Align by default only first pair of columns:
>
  local align = require('mini.align')
  align.setup({
    steps = {
      pre_justify = { align.gen_step.filter('n == 1') }
    },
  })

------------------------------------------------------------------------------
                                                     *MiniAlign.align_strings()*
             `MiniAlign.align_strings`({strings}, {opts}, {steps})
Align strings

For details about alignment process see |MiniAlign-algorithm|.

Parameters~
{strings} `(table)` Array of strings.
{opts} `(table|nil)` Options. Its copy will be passed to steps as second
  argument. Extended with `MiniAlign.config.options`.
  This is a place to control default main steps:
    - `opts.split_pattern` - Lua pattern(s) used to make split parts.
    - `opts.split_exclude_patterns` - which split matches should be ignored.
    - `opts.justify_side` - which direction(s) alignment should be done.
    - `opts.justify_offsets` - offsets tweaking width of first column
    - `opts.merge_delimiter` - which delimiter(s) to use when merging.
  For more information see |MiniAlign.gen_step| entry for corresponding
  default step.
{steps} `(table|nil)` Steps. Extended with `MiniAlign.config.steps`.
  Possible `nil` values are replaced with corresponding default steps:
  - `split` - |MiniAlign.gen_step.default_split()|.
  - `justify` - |MiniAlign.gen_step.default_justify()|.
  - `merge` - |MiniAlign.gen_step.default_merge()|.

------------------------------------------------------------------------------
                                                        *MiniAlign.align_user()*
                         `MiniAlign.align_user`({mode})
Align current region with user-supplied steps

Mostly designed to be used inside mappings.

Will use |MiniAlign.align_strings()| and set the following options in `opts`:
- `justify_offsets` - array of offsets used to achieve actual alignment of
  a region. It is non-trivial (not array of zeros) only for charwise
  selection: offset of first string is computed as width of prefix to the
  left of region start.
- `region` - current affected region (see |MiniAlign-glossary|). Can be
  used to create more advanced steps.
- `mode` - mode of selection (see |MiniAlign-glossary|).

Parameters~
{mode} `(string)` Selection mode. One of "char", "line", "block".

------------------------------------------------------------------------------
                                                     *MiniAlign.action_normal()*
                   `MiniAlign.action_normal`({with_preview})
Perfrom action in Normal mode

Used in Normal mode mapping. No need to use it directly.

Parameters~
{with_preview} `(boolean|nil)` Whether to align with live preview.

------------------------------------------------------------------------------
                                                     *MiniAlign.action_visual()*
                   `MiniAlign.action_visual`({with_preview})
Perfrom action in Visual mode

Used in Visual mode mapping. No need to use it directly.

Parameters~
{with_preview} `(boolean|nil)` Whether to align with live preview.

------------------------------------------------------------------------------
                                                          *MiniAlign.as_parts()*
                         `MiniAlign.as_parts`({arr2d})
Convert 2d array of strings to parts

This function verifies if input is a proper 2d array of strings and adds
methods to its copy.

Class~
{parts}

Fields~
{apply} `(function)` Takes callable `f` and applies it to every part.
  Callable should have signature `(s, data)`: `s` is a string part,
  `data` - table with its data (<row> has row number, <col> has column number).
  Returns new 2d array.

{apply_inplace} `(function)` Takes callable `f` and applies it to every part.
  Should have same signature as in `apply` method. Outputs (should all be
  strings) are assigned in place to a corresponding parts element. Returns
  parts itself to enable method chaining.

{get_dims} `(function)` Return dimensions of parts array: a table with
  <row> and <col> keys having number of rows and number of columns (maximum
  number of elements across all rows).

{group} `(function)` Concatenate neighboring strings based on supplied
  boolean mask and direction (one of "left", default, or "right"). Has
  signature `(mask, direction)` and modifies parts in place. Returns parts
  itself to enable method chaining.
  Example:
  - Parts: { { "a", "b", "c" }, { "d", "e" }, { "f" } }
  - Mask: { { false, false, true }, { true, false }, { false } }
  - Result for direction "left":  { { "abc" },    { "d", "e" }, { "f" } }
  - Result for direction "right": { { "ab","c" }, { "de" },     { "f" } }

{pair} `(function)` Concatenate neighboring element pairs. Takes
  `direction` as input (one of "left", default, or "right") and applies
  `group()` for an alternating mask.
  Example:
  - Parts: { { "a", "b", "c" }, { "d", "e" }, { "f" } }
  - Result for direction "left":  { { "ab", "c" }, { "de" }, { "f" } }
  - Result for direction "right": { { "a", "bc" }, { "de" }, { "f" } }

{slice_col} `(function)` Return column with input index `j`. Note: it might
  not be an array if rows have unequal number of columns.

{slice_row} `(function)` Return row with input index `i`.

{trim} `(function)` Trim elements whitespace. Has signature `(direction, indent)`
  and modifies parts in place. Returns parts itself to enable method chaining.
  - Possible values of `direction`: "both" (default), "left", "right",
  "none". Defines from which side whitespaces should be removed.
  - Possible values of `indent`: "keep" (default), "low", "high", "remove".
  Defines what to do with possible indent (left whitespace of first string
  in a row). Value "keep" keeps it; "low" makes all indent equal to the
  lowest across rows; "high" - highest across rows; "remove" - removes indent.

Usage~
>
  parts = MiniAlign.as_parts({ { 'a', 'b' }, { 'c' } })
  print(vim.inspect(parts.get_dims())) -- Should be { row = 2, col = 2 }

  parts.apply_inplace(function(s, data)
    return ' ' .. data.row .. s .. data.col .. ' '
  end)
  print(vim.inspect(parts)) -- Should be { { ' 1a1 ', ' 1b2 ' }, { ' 2c1 ' } }

  parts.trim('both', 'remove').pair()
  print(vim.inspect(parts)) -- Should be { { '1a11b2' }, { '2c1' } }

------------------------------------------------------------------------------
                                                          *MiniAlign.new_step()*
                     `MiniAlign.new_step`({name}, {action})
Create step

A step is basically a named callable object. Having a name bundled with
some action powers helper status message during interactive alignment process.

Parameters~
{name} `(string)` Step name.
{action} `(function|table)` Step action. Should be a callable object
  (see |vim.is_callable()|).

Return~
`(table)` A table with keys: <name> with `name` argument, <action> with `action`.

------------------------------------------------------------------------------
                                                            *MiniAlign.gen_step*
                              `MiniAlign.gen_step`
Generate common action steps

This is a table with function elements. Call to actually get step.

Each step action is a function that has signature `(object, opts)`, where
`object` is either parts or array of strings (depends on which stage of
alignment process it is assumed to be applied) and `opts` is table of options.

Outputs of elements named `default_*` are used as default corresponding main
step (split, justify, merge). Behavior of all of them depend on values from
supplied options (second argument).

Outputs of other elements depend on both step generator input values and
options supplied at execution. This design is mostly because their output
can be used several times in pre-steps.

Usage~
>
  local align = require('mini.align')
  align.setup({
    modifiers = {
      -- Use 'T' modifier to remove both whitespace and indent
      T = function(steps, _)
        table.insert(steps.pre_justify, align.gen_step.trim('both', 'remove'))
      end,
    },
    options = {
      -- By default align "right", "left", "right", "left", ...
      justify_side = { 'right', 'left' },
    },
    steps = {
      -- Align by default only first pair of columns
      pre_justify = { align.gen_step.filter('n == 1') },
    },
  })

------------------------------------------------------------------------------
                                            *MiniAlign.gen_step.default_split()*
                      `MiniAlign.gen_step.default_split`()
Generate default split step

Output splits strings using matches of Lua pattern(s) from `split_pattern`
option which are not dismissed by `split_exclude_patterns` option.

Outline of how single string is split:
- Convert `split_pattern` option to array of strings (string is converted
  as one-element array). This array will be recycled in case there are more
  split matches than in converted `split_pattern` array (which almost always).
- Find all forbidden spans (intervals inside string) - all matches of all
  patterns in `split_exclude_patterns`.
- Find match for the next pattern. If it is not inside any forbidden span,
  add preceding unmatched substring and matched split as two parts. Repeat
  with the next pattern.
- If no pattern match is found, add the rest of string as final part.

Output uses following options (as part second argument, `opts` table):
- <split_pattern> - string or array of strings used to detect split matches
  and create parts. Default: `''` meaning no matches (whole string is used
  as part). Examples: `'%s+'`, `{ '<', '>' }`.
- <split_exclude_patterns> - array of strings defining which regions to
  exclude from being matched. Default: `{}`. Examples: `{ '".-"', '^%s*#.*' }`.

Return~
`(table)` A step named "split" and with appropriate callable action.

See also~
|MiniAlign.gen_step.ignore_split()| heavily uses `split_exclude_patterns`.

------------------------------------------------------------------------------
                                          *MiniAlign.gen_step.default_justify()*
                     `MiniAlign.gen_step.default_justify`()
Generate default justify step

Output makes column elements of string parts have equal width by adding
left and/or right whitespace padding. Which side(s) to pad is defined by
`justify_side` option. Width of first column can be tweaked with `justify_offsets`
option.

Outline of how parts are justified:
- Convert `justify_side` option to array of strings (single string is
  converted as one-element array). Recycle this array to have length equal
  to number of columns in parts.
- For all columns compute maximum width of strings from it (add offsets from
  `justify_offsets` to first column widths). Note: for left alignment, width
  of last row element does not affect column width. This is mainly because
  it won't be padded and helps dealing with "no single match" lines.
- Make all elements have same width inside column by adding appropriate
  amount of whitespace. Which side(s) to add is controlled by the corresponding
  `justify_side` array element. Note: padding is done with spaces which
  might conflict with tab indentation.

Output uses following options (as part second argument, `opts` table):
- <justify_side> - string or array of strings. Each element can be one of
  "left" (pad right side), "center" (pad both sides equally), "right" (pad
  left side), "none" (no padding). Default: "left".
- <justify_offsets> - array of numeric left offsets of rows. Used to adjust
  for possible not equal indents, like in case of charwise selection when
  left edge is not on the first column. Default: array of zeros. Set
  automatically during interactive alignment in charwise mode.

Return~
`(table)` A step named "justify" and with appropriate callable action.

------------------------------------------------------------------------------
                                            *MiniAlign.gen_step.default_merge()*
                      `MiniAlign.gen_step.default_merge`()
Generate default merge step

Output merges rows of parts into strings by placing merge delimiter(s)
between them.

Outline of how parts are converted to array of strings:
- Convert `merge_delimiter` option to array of strings (single string is
  converted as one-element array). Recycle this array to have length equal
  to number of columns in parts minus 1.
- Exclude empty strings from parts. They add nothing to output except extra
  usage of merge delimiter.
- Concatenate each row interleaving with array of merge delimiters.

Output uses following options (as part second argument, `opts` table):
- <merge_delimiter> - string or array of strings. Default: `''`.
  Examples: `' '`, `{ '', ' ' }`.

Return~
`(table)` A step named "merge" and with appropriate callable action.

------------------------------------------------------------------------------
                                                   *MiniAlign.gen_step.filter()*
                      `MiniAlign.gen_step.filter`({expr})
Generate filter step

Construct function predicate from supplied Lua string expression and make
step evaluating it on every part element.

Outline of how filtering is done:
- Convert Lua filtering expression into function predicate which can be
  evaluated in manually created context (some specific variables being set).
- Compute boolean mask for parts by applying predicate to each element of
  2d array with special variables set to specific values (see next section).
- Group parts with compted mask. See `group()` method of parts in
  |MiniAlign.as_parts()|.

Special variables which can be used in expression:
- <row> - row number of current element.
- <ROW> - total number of rows in parts.
- <col> - column number of current element.
- <COL> - total number of columns in current row.
- <s>   - string value of current element.
- <n>   - column pair number of current element. Useful when filtering by
          result of pattern splitting.
- <N>   - total number of column pairs in current row.
- All variables from global table `_G`.

Tips:
- This general filtering approach can be used to both include and exclude
  certain parts from alignment. Examples:
    - Use `row ~= 2` to align all parts except from second row.
    - Use `n == 1` to align only by first pair of columns.
- Filtering by last equal sign usually can be done with `n >= (N - 1)`
  (because there is usually something to the right of it).

Parameters~
{expr} `(string)` Lua expression as a string which will be used as predicate.

Return~
`(table)` A step named "filter" and with appropriate callable action.

------------------------------------------------------------------------------
                                             *MiniAlign.gen_step.ignore_split()*
        `MiniAlign.gen_step.ignore_split`({patterns}, {exclude_comment})
Generate ignore step

Output adds certain values to `split_exclude_patterns` option. Should be
used as pre-split step.

Parameters~
{patterns} `(table)` Array of patterns to be added to
  `split_exclude_patterns` as is. Default: `{ [[".-"]] }` (excludes strings
  for most cases).
{exclude_comment} `(boolean)` Whether to add comment pattern to
  `split_exclude_patterns`. Comment pattern is derived from 'commentstring'
  option. Default: `true`.

Return~
`(table)` A step named "ignore" and with appropriate callable action.

See also~
|MiniAlign.gen_step.default_split()| for details about
  `split_exclude_patterns` option.

------------------------------------------------------------------------------
                                                     *MiniAlign.gen_step.pair()*
                     `MiniAlign.gen_step.pair`({direction})
Generate pair step

Output calls `pair()` method of parts (see |MiniAlign.as_parts()|) with
supplied `direction` argument.

Parameters~
{direction} `(string)` Which direction to pair. One of "left" (default) or


Return~
`(table)` A step named "pair" and with appropriate callable action.

------------------------------------------------------------------------------
                                                     *MiniAlign.gen_step.trim()*
                `MiniAlign.gen_step.trim`({direction}, {indent})
Generate trim step

Output calls `trim()` method of parts (see |MiniAlign.as_parts()|) with
supplied `direction` and `indent` arguments.

Parameters~
{direction} `(string)` Which sides to trim whitespace. One of "both"
  (default), "left", "right", "none".
{indent} `(string)` What to do with possible indent (left whitespace of first
  string in a row). One of "keep" (default), "low", "high", "remove".

Return~
`(table)` A step named "trim" and with appropriate callable action.


 vim:tw=78:ts=8:noet:ft=help:norl: