index

`"a2d<C-V>3gE`: Vim normal mode grammar

This is a purely academic exercise in doing a proper linguistic description of Vim’s normal mode grammar. Read on if you’re feeling adventurous!

Much has been said of “nouns” and “verbs” in Vim’s normal mode command language. Unfortunately, a proper linguistic account has not been forthcoming. In this article I will attempt to give such an account.

Grammar outline

This grammar outline focuses on the syntax of Vim grammar. We will examine all the constituents in the maximal grammatical sentence given in the title, "a2d<C-V>3gE. We will walk through the syntax first, covering all the constituents comprehensively, then talk about word order, word classes, and some morphological phenomena in separate sections.

Syntax overview

Intransitive sentence

The minimal sentence in Vim consists of just a verb.

p “put”
$ “move to the end of the line”

These are Vim’s intransitive sentences. Only a few intransitive verbs like J p ~ are capable of this usage. Also capable of functioning as verbs of intransitive sentences are all the motion verbs (motions) w ^ gE etc. Motion verbs have a number of properties which set them apart from the other intransitive verbs. We will see more of them in the next and following sections.

Since all of Vim’s normal mode sentences are commands, we will take them as imperatives: “put!”, “move!”. A subject is never expressed.

Transitive sentence

Some verbs take an object argument to form basic transitive sentences.

yaw “yank an entire word”

Transitive verbs (operators) include y d gu. An object can be one of two kinds: a common noun (text object) as in aw “an entire word” and i( “the material inside the brackets”; and a motion verb in argument function.

Here is an example for a motion verb in argument function:

dw “delete to the beginning of the next word”

Used as an argument, a motion verb comes to denote the part that would be covered when executing the motion. So, w as a verb means “move the cursor to the beginning of the next word”, but w as an object means “the part covered by moving from the current cursor position to the beginning of the next word”.

A special property of a motion verb, its motion type, becomes apparent when the motion verb is in argument function: some motion verbs yield a linewise object whereas others yield a characterwise one.

Quantifiers

Both the verb and the object can be quantified by preceding them with a quantifier (count):

5gUe “five times upcase the text covered by moving the cursor to the end of the next word”
c2k “change the lines covered by moving the cursor up two lines”.

Both quantifiers can be given at the same time, as in our title sentence.

2d3gE “twice delete to the end of three whitespace-separated words backwards”

The verbal and the object quantifier form an intimate relationship. They are multiplied and their product quantifies the whole predication. In fact, Vim does not remember the quantifiers individually, only the product. When the . command is used with a count the entire predicate quantification is replaced with the new count. It is therefore not implausible to argue that the verb and object quantifiers represent one discontinuous syntactical constituent.

We will call the constituents covered so far – verb, object, and optional quantifiers – the “core predication”. For our title sentence the core predication is 2d3gE.

“Goal”

Any core predication can be preceded by a Goal¹ (register). The register specifies where the result of the predication should go or where it should draw from.

"adE “into register a delete what’s covered by moving to the end of the next whitespace-separated word”
"xp “put what’s in register x”

When no register is given, the default register " is understood.

It should be mentioned that the default register isn’t the only thing that is understood. Place and time – the current cursor position and “now” – are also part of the default context, as is a count of 1. Thus, the simplest sentence p actually means “put (it) (here) (once)”, i.e. ""1p. “Now” is always implicit, all sentences are immediate.

Adverb

A less well known part of Vim’s grammar is the adverb (called “motion force” in the source; the Vim documentation doesn’t have a specific term for this). There are three adverbs v V <C-V>, which specify characterwise, linewise, and blockwise operation. An adverb is only acceptable in transitive sentences. When present, it changes the way in which the verb applies to the object.

Consider these sentences:

g?} “rot-13-encode (characterwise) to the end of the paragraph”
g?V} “rot-13-encode linewise to the end of the paragraph”

Both sentences make the same simple predication, but the second has the linewise V adverb. The motion verb } has characterwise motion type when used as an argument. Its meaning is “the part covered by moving from the cursor position to the end of the paragraph”.

With the adverb V the “part covered” is to apply linewise, so the command g?V} will rot-13-encode all lines from whatever line the cursor was on to the line at the end of the paragraph. The adverb has essentially changed how the core predication must be understood.

An alternative analysis for v V <C-V> could seem possible. Since their ultimate effect is to modify the extent of the text operated upon we might view them as adjectives, modifiers of the object.

There is evidence against that analysis: the dependence of the adverbs on the verb, the fact that they can combine with Ex command movement, and their relative syntactic independence from the object (d2vj ~ dv2j) provide clues. The decisive evidence, however, comes from their interplay with the quantifiers. 2yV3w, executed on the buffer shown below, yanks three lines, not four: the adverb has scope over the whole core predication with both verbal and object quantifiers. Its scope is not limited to just the quantified object.

[a]b cd
ef gh
ij kl mn
op qr

This concludes our overview of the syntax. We return to our title sentence "a2d<C-V>3gE in the next section.

Word order

Typologically speaking, Vim’s normal mode syntax is VO with fixed word order (slots). The only exception to the fixed word order is in the order of adverb and object quantifier: these can be ordered freely with no difference in meaning.²

As explained above, the verbal and the object quantifier have an intimate relationship and may be considered one discontinuous constituent.

Here are examples and syntax schemas for intransitive and transitive sentences.

Intransitive sentence	[`"x`]	[`9`]	`gp`
	register	count	"command"/motion
	Goal	quantifier	intransitive/motion verb

Transitive sentence	[`"a`]	[`2`]	`d`	[`<C-V>`] ~ [`3`]	`gE`
	register	count	operator	"force" ~ count	motion/text object
	Goal	quantifier	transitive verb	adverb ~ quantifier	object

So, coming back to the title sentence "a2d<C-V>3gE, we may call this a maximal transitive sentence in Vim’s normal mode language. In a maximal sentence every constituent slot is filled.

Even in this situation the language retains its human-language character, and we may translate approximately into English: “Into register a twice delete blockwise to the end of three whitespace-separated words backwards”.

Word classes

We are now well prepared to identify Vim’s parts of speech, or word classes.

The major word classes are large. They are extensible through mappings and user-provided language elements. They are referential in that they refer to actions on the text, or to parts of the text.

As major word classes we can identify the following:

intransitive verbs, e.g. p J ~
transitive verbs, e.g. y d gu
motion verbs, e.g. G w 0 ;
- complex motion verbs, { f F t T }
common nouns, e.g. iw as a[

The minor word classes are small and closed, that is not extensible. Its members are more or less enumerable. The minor word classes are:

quantifiers, { 1, 2, 3, … }
Goal, { [-a-zA-Z0-9"*+~_/] }
adverb, { v, V, <C-V> }

Morphological phenomena

We are not primarily concerned with form in this article, so a few remarks on the morphology of Vim’s words should suffice. The form of verbs and objects is generally guided by some kind of mnemotechnic principle: they are supposed to be easy to remember. Most verbs are just one character in size, but there are exceptions like gp (intransitive verb), g? (transitive verb), gE (motion verb). Common nouns are two characters long, and start with a or i.

There is one interesting phenomenon touching on both morphology and syntax. Consider these sentences:

dd “delete one line”
cc “change one line”
yy “yank one line”

These are exactly equivalent to d_ c_ y_. These last sentences are perfectly regular transitive sentences: verb plus the object _ “the current line”. They can be extended with optional constituents, e.g. "xd4_ “into register x delete four lines under the cursor”.

Now, the object _ is actually a rather obscure motion verb meaning “move count−1 lines downward”. So, to make a very common utterance like d_ easy to type, the object has been assimilated to the verb: dd. But despite the change in form, the second “d” remains an ordinary object, so it is possible to say dv3d meaning “delete characterwise three lines under the cursor”. Since this is not well known, dd is often interpreted as a single intransitive verb by users and is firmly idiomatic. What we are dealing with here is an instance of grammaticalization.

A final point are intransitive verbs with internalized objects:

D “delete to the end of the line”
x “delete the character under the cursor”

These are synonymous with the verb-object predicates d$ and dl. Internally, Vim does in fact translate them to their verb-object equivalent, and . repeats the verb-object sentence.

The interesting case of S takes part in both morphological phenomena shown here: it is translated to cc which in turn is the grammaticalized version of c_.

Outlook

Vim’s normal mode grammar, the topic of the above grammar outline, is the part of Vim that lends itself best to a description in linguistic terms. It is the part that most directly gives an impression of natural language.

But Vim is much bigger than that. There are commands – intransitive verbs – which switch into a different but related grammar:

v “go to Visual mode and select the character under the cursor”

After selecting something in Visual mode and pressing an operator – a transitive verb – the verb is applied to the selection, the object: verb and object have switched places to OV.

A clause of an entirely different grammar can even be embedded as the object of a transitive sentence. Consider this single sentence:

d:call setpos(".",[0,3,4,0])<Enter> “delete to line 3, column 4”

In this example, the object starts with : which makes it possible to use Ex commands to specify the object. The Ex language has its own, unrelated grammar.

Finally, since to be frank we are not dealing with an actual natural language, there are a fair number of verbs which do not fit well in a language interpretation in the terms used in this outline: q “record keystrokes”, m “drop a mark”, & “repeat last substitution”, and so on.

"a2d<C-V>3gE: Vim normal mode grammar