Skip to content

Latest commit

History

History
1127 lines (730 loc) 路 25.1 KB

Spec Additions.md

File metadata and controls

1127 lines (730 loc) 路 25.1 KB

Spec Additions

Spec Markdown makes some additions to Markdown to support cases relevant to writing technical specs and documentation. It attempts to be as minimally invasive as possible, leveraging existing Markdown formatting features whenever possible so Spec Markdown documents may render adequately as regular Markdown.

Spec Markdown also makes restrictions to the overall format of the Markdown document in order to derive a structure to the entire document.

Link Anything

Everything unique in a Spec Markdown file has a link created for it. Sections each have a link, as do named Algorithms and Grammar. You'll find that Notes and Examples are also given stable links based on their contents, just in case things move around.

However, you can also link anything in a Spec Markdown file. Just highlight any bit of text and a link will be created just for that selection, making referencing specific parts of your document easy. Try it here!

Title and Introduction

A Spec Markdown document must start with a header which will be used as the title of the document. Any content between this and the next header will become the introduction to the document.

A Spec Markdown document starts in this form:

# Spec Markdown

Introductory paragraph.

# First Section Header

Note: For backwards-compatibility, a setext style header can be used for a document title.

Sections

A Spec Markdown document is separated into a sequence and hierarchy of sections. Those sections can then be used as navigation points and can be used to create a table of contents. A section is started by a header and ends at either the next header of similar or greater precedence or the end of the document. A section can contain other sections if their headers are of lower precedence.

Section Headers

Regular Markdown supports two styles of headers, Setext and atx, however Spec Markdown only supports atx style headers as section headers.

# Header

Only use Setext style headers for the title of the document.

Header
------

Subsection Headers

While sections are numbered and appear in the table of contents, a subsection is similar but not numbered or in the table of contents.

This is a subsection

The subsection's content appears below the subsection header.

Another subsection

Sections may contain multiple subsections, but subsections cannot contain sections or subsections.

Table of Contents

A table of contents is automatically generated from the hierarchy of sections in the Spec Markdown document.

Section Numbers

A number is associated with each section, starting with 1. In a hierarchy of sections, the parent sections are joined with dots. This provides an unambiguous location identifier for a given section in a document.

You can specify these section numbers directly in your Markdown documents if you wish by writing them directly after the # and before the text of the header.

3.2.3.8. Custom Numbers

If the section number is written in the document, the last number will be used as the number for that section. This is useful when writing a proposal against an existing spec and wish to reference a particular section.

The header for this section was written as

#### 3.2.3.8. Custom Numbers

Appendix / Annex Sections

If a top level section is written with a letter, such as A instead of a number, that will begin an Appendix section.

# A. Appendix: Grammar

Smart Characters

The Spec Markdown renderer will replace easy to type characters like quotes and dashes with their appropriate typographic entities. These replacements will not occur within blocks of code.

Quotes and Dashes

Prose text has "smart quotes", hyphens, en-dashes and em-dashes--you shouldn't have to think about it, they'll just work.

For example, a quote of a quote (with an inner apostrophe and emphasis for flair):

"She told me that 'he isn't here right *now*' - so I left."

Will render as:

"She told me that 'he isn't here right now' - so I left."

Escaped \"quotes \'and single\-quotes\'\" becomes: "quotes 'and single-quotes'".

Math

Math operators like >=, <=, and ~= can be written as >=, <=, and ~=.

Escaped \>= \<= \~= becomes: >= <= ~=.

Arrows

Smart arrows -> and <- and <-> can be written as ->, <- and <->.

Fat smart arrows => and <== and <=> can be written as =>, <== and <=>.

Escaped \-> \<- \<-> \=> \<== \<=> becomes: -> <- <-> => <== <=>.

Tables

Similar to Github flavored Markdown

| This | is a | table |
| ---- | ---: | :---: |
| key  | val  | etc   |

Produces the following:

This is a table
key val etc

Table cells can contain any content that a paragraph can contain.

Comments

Spec Markdown will ignore HTML style comments before or within blocks of text. By default these comments are omitted from the resulting HTML, but can be included by passing the option --includeComments when calling spec-md.

For example this paragraph:

This is a <!-- comment inside of a --> paragraph.

Produces the following:

This is a paragraph.

Definitions

Spec Markdown provides two forms for defining terms, definition lists and definition paragraphs.

Definition List

A definition list is written as a defined term on a single line followed by one or more definition lines, each starting with a : .

Cookie
:   A small piece of data that a server sends to the user's web browser. The
    browser may store it and send it back with later requests to the same server.
:   A delicious snack, often containing chocolate chips.

Produces the following:

Cookie : A small piece of data that a server sends to the user's web browser. The browser may store it and send it back with later requests to the same server. : A delicious snack, often containing chocolate chips.

Note: Term definitions can optionally be separated by a single blank line.

Definition Paragraph

A definition paragraph starts with a :: and contains an italicized term. This is useful when it is easier to define a term in a sentence containing that term.

:: The study of *Philosophy* investigates general and fundamental questions.

Produces the following:

:: The study of Philosophy investigates general and fundamental questions.

Definition References

A defined term can be later referenced by italicizing that term. Referenced terms are case insensitive.

After studying *philosophy*, you may eat a *cookie*.

Produces the following:

After studying philosophy, you may eat a cookie.

Note

Notes can be written inline with a spec document, and are often helpful to supply non-normative explanatory text or caveats in a differently formatted style. Case insensitive, the : is optional.

Notes automatically have short links generated for them. If the contents of the note changes, so will the link URL. However if a note moves around, or content around the note changes the existing links will still point to the right place, very useful for consistently evolving specifications!

Note: Notes are awesome.

Produces the following:

Note: Notes are awesome.

Todo

It's often helpful to write a draft of a document and leave "to-do" comments in not-yet-completed sections. Case insensitive, the : is optional.

TODO: finish this section

Produces the following:

TODO: finish this section

Note: You can also write TK in place of TODO, nerds.

Syntax Highlighting

Spec Markdown will apply syntax highlighting to blocks of code if a github-flavored-markdown style language is supplied.

You may provide a highlight function as an option to customize this behavior.

To render this highlighted javascript:

```js
const baz = foo("bar");
```

Produces the following:

const baz = foo("bar");

You may also prefix your highlight function with "raw" if you want to avoid other tools, such as Prettier, from interpreting a code block.

```raw js
const baz = foo("bar");
```

Produces the following:

const baz = foo("bar");

Examples

Spec Markdown helps you write examples, visually indicaticating the difference from normative code blocks, and generating permalinks to those examples. Just write example after the ```.

```example
const great = useOf.example("code");
```

Produces the following:

const great = useOf.example("code");

Examples can also be syntax highlighted, by placing the language directly before writing example:

```js example
const great = useOf.example("code");
```

Produces the following:

const great = useOf.example("code");

Counter Examples

In addition to examples, Spec Markdown helps you write counter-examples, which are examples of things you should not do. These are visually indicated as different from normative code blocks and other examples. Just write counter-example after the ``` (and optional language).

```js counter-example
const shit = dontSwear();
```

Produces the following:

const shit = dontSwear();

Imports

When compiled, an import reference will be inlined into the same document. An import reference looks like a link to a ".md" file as a single paragraph.

[AnythingGoesHere](SomeName.md)

You can optionally prefix the import reference with # characters to describe at what section level the import should apply. By default an import reference will be imported as a child of the current section.

Inline editing

A portion of the CriticMarkup spec is supported.

For example, we can {++add++} or {--remove--} text with the {++add++} or {--remove--} syntax.

Block editing

We can also add and remove entire blocks of content, by using {++ or {-- on their own line with empty lines on either side:

{++

These paragraphs

have been added.

++}

And

{--

These paragraphs

have been removed.

--}

By typing:

{++

These paragraphs

have been *added*.

++}

And

{--

These paragraphs

have been *removed*.

--}

Note: imports and section headers cannot be included in a added or removed section to preserve the ability to render a table of contents.

Algorithms

Specifications for procedures or algorithms can be defined in terms of nested markdown lists. These lists can be of any kind, but will always have ordered formatting. The bullet labeling for algorithms is specific will cycle between decimal, lower-alpha, and lower-roman.

An algorithm definition also describes its arguments in terms of variables.

Algorithm(arg) :
  1. first
  1. then
    * substep
      * deeper substep
      * another deep substep
    * another step
  1. okay

Produces the following:

Algorithm(arg) :

  1. first
  2. then * substep
    • deeper substep
    • another deep substep * another step
  3. okay

Grammar

Spec Markdown makes it easier to describe context-free grammatical productions.

Grammars are defined by a sequence of terminal characters or sequence of characters, which are then referenced by non-terminal rules. The definition of a non-terminal is referred to as a production.

Grammar Production

The : token indicates an "is defined as" production for a non-terminal, where a single definition can be written directly after the :.

PBJ : Bread PeanutButter Jelly Bread

Produces the following:

PBJ : Bread PeanutButter Jelly Bread

Or if {PBJ} has definition options, they are written immediately after as a Markdown list.

PBJ :
  - Bread PeanutButter Jelly Bread
  - Bread Jelly PeanutButter Bread

Produces the following:

PBJ :

  • Bread PeanutButter Jelly Bread
  • Bread Jelly PeanutButter Bread

Each definition is a space seperated list of terminal or non-terminal tokens, and may also include conditionals and constraints.

Definition lists aren't required to be indented:

PBJ :

- Bread PeanutButter Jelly Bread
- Bread Jelly PeanutButter Bread

Produces the following:

PBJ :

  • Bread PeanutButter Jelly Bread
  • Bread Jelly PeanutButter Bread

Production types

Often languages wish to specify different types of grammar productions, such as lexical or syntactical, or if certain characters line whitespace or newlines are permitted between symbols in the right-hand-side. Spec-md allows this this distinction based on the number of colons:

TypeOne : `type` `one`

TypeTwo :: `type` `two`

TypeThree ::: `type` `three`

Produces the following:

TypeOne : type one

TypeTwo :: type two

TypeThree ::: type three

One of

If each definition option is a single token, it can be expressed as a "one of" expression instead of a markdown list.

AssignmentOperator : one of *= `/=` %= += -= <<= >>= >>>= &= ^= |=

Produces the following:

AssignmentOperator : one of *= /= %= += -= <<= >>= >>>= &= ^= |=

"one of" can also be followed by a line break and multiple lines of tokens. To improve legibility in other tools, each line may optionally begin with a bullet.

Keyword : one of
  - break     do        in          typeof
  - case      else      instanceof  var
  - catch     export    new         void
  - class     extends   return      while
  - const     finally   super       with
  - continue  for       switch      yield
  - debugger  function  this
  - default   if        throw
  - delete    import    try

Produces the following:

Keyword : one of

  • break do in typeof
  • case else instanceof var
  • catch export new void
  • class extends return while
  • const finally super with
  • continue for switch yield
  • debugger function this
  • default if throw
  • delete import try

Non Terminal Token

Non-terminal tokens with a defined as a grammar production can be referred to in other grammar productions. Non-terminals must match the regular expression {/[A-Z][_a-zA-Z]*/}. That is, they must start with an uppercase letter, followed by any number of letters or underscores.

Prose

Grammars can describe arbitrary rules by using prose within a grammar definition by using "quotes".

Sandwich : Bread "Any kind of topping" Bread

Produces the following:

Sandwich : Bread "Any kind of topping" Bread

Terminal Token

Terminal tokens refer to a character or sequence of characters. They can be written unadorned in the grammar definition.

BalancedParens : ( BalancedParens )

Produces the following:

BalancedParens : ( BalancedParens )

Any sequence of characters can be written to indicate a terminal token:

WhileStatement : while ( Expression ) { Statements }

Produces

WhileStatement : while ( Expression ) { Statements }

Terminals can also be quoted with back-ticks ` to remove any ambiguity from other meanings, for example to allow a terminal token to start with an uppercase letter, or a slash / or backslash \, or later contain a ] or }.

DivisionExpression : Expression `/` Expression

Produces

DivisionExpression : Expression / Expression

Regular Expression

When a grammar is intended to be interpretted as a single token and can be clearly written as a regular expression, you can do so directly.

UppercaseWord : /[A-Z][a-z]*/

Produces the following:

UppercaseWord : /[A-Z][a-z]*/

Quantifiers

Tokens can be followed by quantifiers to alter their meaning and as a short-hand for common patterns of optionality and repetition.

Optional Tokens

A subscript suffix Token? renders as {Token?} and is a shorthand for two possible definitions, one including that token and one excluding it.

Sentence : Noun Verb Adverb?

Produces the following:

Sentence : Noun Verb Adverb?

Which is shorthand for:

Sentence :

  • Noun Verb Adverb
  • Noun Verb

Token Lists

A subscript suffix Token+ renders as {Token+} and is shorthand for a list of one or more of that token.

Book : Cover Page+ Cover

Produces the following:

Book : Cover Page+ Cover

Which, unless your specification document declares otherwise, is shorthand for:

Book : Cover Page_list Cover

Page_list :

  • Page_list Page
  • Page

Some specifications may wish to declare {Token+} as a shorthand for a comma-separated list, in which case the previous example would be shorthand for:

Book : Cover Page_list Cover

Page_list :

  • Page_list , Page
  • Page

Optional Lists

A subscript suffix Token* renders as {Token*} and is shorthand for an optional list, which describes zero or more of that token.

Sandwich : Bread Topping* Bread

Produces the following:

Sandwich : Bread Topping* Bread

Which is shorthand for:

Sandwich :

  • Bread Topping_list Bread
  • Bread Bread

Topping_list :

  • Topping_list Topping
  • Topping

Use with Non-Terminals

Quantifiers also apply to non-terminal tokens with the same rules. For example:

UnionMembers :
  - UnionMembers | NamedType
  - `|`? NamedType

Produces the following:

UnionMembers :

  • UnionMembers | NamedType
  • |? NamedType

However, unquoted non-terminals may use the *, ? and + characters, so always quote the terminal if the intent is to apply a quantifer.

UnionMembers :
  - UnionMembers | NamedType
  - |? NamedType

Produces the terminal |?, not an optional |:

UnionMembers :

  • UnionMembers | NamedType
  • |? NamedType

Conditional Parameters

It can be a useful short-hand to provide conditional parameters when defining a non-terminal token rather than defining two very similar non-terminals.

A conditional parameter is written in braces Token[Param] and renders as {Token[Param]}. When used in definitions is shorthand for two symbol definitions: one appended with that parameter name, the other without.

Example[WithCondition] : "Definition TBD"

Produces the following:

Example[WithCondition] : "Definition TBD"

Which is shorthand for:

Example : "Definition TBD"

Example_WithCondition : "Definition TBD"

The conditions are applied at the beginning of a definition for the non-terminal by prefixing with [if Param] (alternatively [+Param]) or [if not Param] (alternatively [~Param]) to only include the definition when the variant with the conditional parameter is or is not used, respectively.

Example[WithCondition] :
  - A
  - [if WithCondition] B
  - [if not WithCondition] C
  - [+WithCondition] D
  - [~WithCondition] E

Produces the following:

Example[WithCondition] :

  • A
  • [if WithCondition] B
  • [if not WithCondition] C
  • [+WithCondition] D
  • [~WithCondition] E

Which is shorthand for:

Example :

  • A
  • C
  • E

Example_WithCondition :

  • A
  • B
  • D

The same bracket suffix on a non-terminal within a rule is shorthand for using that variant of the rule. If the parameter starts with ?, that form of the symbol is conditionally used only in the derived production with the same parameter. If the parameter starts with !, that form of the symbol is only used when in the derived production without that parameter.

Example[WithCondition] :
  - Example
  - Example[WithCondition]
  - Example[?WithCondition]
  - Example[!WithCondition]

Produces the following:

Example[WithCondition] :

  • Example
  • Example[WithCondition]
  • Example[?WithCondition]
  • Example[!WithCondition]

Which is shorthand for:

Example :

  • Example
  • Example_WithCondition
  • Example
  • Example_WithCondition

Example_WithCondition :

  • Example
  • Example_WithCondition
  • Example_WithCondition
  • Example

Multiple conditional parameters can be used on both the production definition and on non-terminals within a rule, in which case it is short form for the permutation of all conditions:

Example[P, Q] :
  - [if P] `p`
  - [if Q] `q`
  - Example[!P, ?Q]

Produces the following:

Example[P, Q] :

  • [if P] p
  • [if Q] q
  • Example[!P, ?Q]

Which is shorthand for:

Example :

  • Example_P

Example_P :

  • p
  • Example

Example_Q :

  • q
  • Example_P_Q

Example_P_Q :

  • p
  • q
  • Example_Q

Conditional parameters on a usage can be followed by a quantifier.

Example[P, ?Q]*

Produces the following:

{Example[P, ?Q]*}

Constraints

Any token can be followed by "but not" or "but not one of" to place a further constraint on the previous token:

Example : A B but not foo or bar

Produces the following:

Example : A B but not foo or bar

Optionally can mention "one of", this will be omitted when rendered. Commas can be used instead of "or".

Example : A B but not one of foo, bar

Produces the following:

Example : A B but not one of foo, bar

Meta Tokens

Spec Markdown can specify some tokens which do not consume any characters.

The empty set, written [empty] appears as {[empty]} can be used to define a non-terminal as matching no terminal or non-terminal tokens.

Example : [empty]

Produces the following:

Example : [empty]

Lookaheads can appear anywhere in a sequence of tokens, and describe additional constraints on the following token.

Example :
  - [lookahead token] Token
  - [lookahead ! token] Token
  - [lookahead != token] Token
  - [lookahead NonTerminal] Token
  - [lookahead ! NonTerminal] Token
  - [lookahead != NonTerminal] Token
  - [lookahead {token, set}] Token
  - [lookahead ! {token, set}] Token
  - [lookahead != {token, set}] Token

Produces the following:

Example :

  • [lookahead token] Token
  • [lookahead ! token] Token
  • [lookahead != token] Token
  • [lookahead NonTerminal] Token
  • [lookahead ! NonTerminal] Token
  • [lookahead != NonTerminal] Token
  • [lookahead {token, set}] Token
  • [lookahead ! {token, set}] Token
  • [lookahead != {token, set}] Token

Grammar Semantics

Once grammar is defined, it can be useful to define the semantics of the grammar in terms of algorithm steps. A single grammar definition followed by a list is interpretted as a grammar semantic:

PBJ : Bread PeanutButter Jelly Bread

* Let {bottomBread} be the result of placing the first {Bread} on the plate.
* Let {pbSpread} be the result of getting {PeanutButter} from the jar.
* Spread {pbSpread} onto {bottomBread}.
* Let {topBread} be the result of placing the last {Bread} on the plate.
* Let {jamSpread} be the result of getting {Jelly} from the jar.
* Spread {jamSpread} onto {topBread}.
* Let {sandwich} be the result of rotating {topBread} 180&deg; and placing on {bottomBread}.
* Return {sandwich}.

Produces the following:

PBJ : Bread PeanutButter Jelly Bread

  • Let {bottomBread} be the result of placing the first {Bread} on the plate.
  • Let {pbSpread} be the result of {PeanutButter}.
  • Spread {pbSpread} onto {bottomBread}.
  • Let {topBread} be the result of placing the last {Bread} on the plate.
  • Let {jamSpread} be the result of {Jelly}.
  • Spread {jamSpread} onto {topBread}.
  • Let {sandwich} be the result of rotating {topBread} 180掳 and placing on {bottomBread}.
  • Return {sandwich}.

Value Literals

Value literals allow any text to refer to a value which has semantic meaning in the specification by wrapping it in { } curly brace characters.

I can reference {foo}, {"foo"}, {null}, {true}.

Produces the following:

I can reference {foo}, {"foo"}, {null}, {true}.

Variables

Write {foo} to produce a variable (represented by a <var> tag) like {foo}.

Keywords

Some known keywords like {null}, {undefined}, {true} and {false} are rendered as constants instead of variables.

String literal

Write {"foo"} to produce a string literal like {"foo"}.

Grammar tokens

Any grammar token can be written inline, like {Example} to represent the non-terminal token {Example}, {`terminal`} to represent the terminal token {terminal}. Even meta tokens like {[empty]} for {[empty]} and {[lookahead !{ x, y }]} for {[lookahead !{ x, y }]}.

Algorithm calls

A call to an algorithm can be expressed as a value literal:

{Algorithm(foo, "string", null)}

Produces the following:

{Algorithm(foo, "string", null)}

Biblio

By supplying a "biblio" key in a metadata file, you can have Algorithm calls and Non-terminal tokens which are not defined in this spec to link to where they are defined.

spec-md -m metadata.json myspec.md

Where metadata.json includes:

{
  "biblio": {
    "http://people.mozilla.org/~jorendorff/es6-draft.html": {
      "Identifier": "#sec-names-and-keywords",
      "PrimaryExpression": "#sec-primary-expression",
      "ReturnIfAbrupt()": "#sec-returnifabrupt",
      "Get()": "#sec-get-o-p"
    }
  }
}

Then referring to these tokens will link out to that page.

MemberExpression : PrimaryExpression . Identifier

  * Let {reference} be the result of evaluating {PrimaryExpression}.
  * Let {propName} be the string value of {Identifier}.
  * Let {value} be {Get(reference, propName)}.
  * {ReturnIfAbrupt(value)}.
  * Return {value}.

Produces the following:

MemberExpression : PrimaryExpression . Identifier

  • Let {reference} be the result of evaluating {PrimaryExpression}.
  • Let {propName} be the string value of {Identifier}.
  • Let {value} be {Get(reference, propName)}.
  • {ReturnIfAbrupt(value)}.
  • Return {value}.