Skip to content
Quick links:   Flags   Verbs   Functions   Glossary   Release docs

Differences from other programming languages

The Miller programming language is intended to be straightforward and familiar, as well as not overly complex. It doesn't try to break new ground in terms of syntax; there are no classes or closures, and so on.

While the Principle of Least Surprise is often held to, nonetheless the following may be surprising.

No ++ or --

There is no ++ or -- operator. To increment x, use x = x+1 or x += 1, and similarly for decrement.

Semicolons as delimiters

You don't need a semicolon to end expressions, only to separate them. This was done intentionally from the very start of Miller: you should be able to do simple things like mlr put '$z = $x * $y' myfile.dat without needing a semicolon.

Note that since you also don't need a semicolon before or after closing curly braces (such as begin/end blocks, if-statements, for-loops, etc.) it's easy to key in a few semicolon-free statements, and then to forget a semicolon where one is needed . The parser tries to remind you about semicolons whenever there's a chance a missing semicolon might be involved in a parse error.

mlr --csv --from example.csv put -q '
  begin {
    @count = 0 # No semicolon required -- before closing curly brace
  $x=1         # No semicolon required -- at end of expression
mlr --csv --from example.csv put -q '
  begin {
    @count = 0 # No semicolon required -- before closing curly brace
  $x=1         # Needs a semicolon after it
  $y=2         # No semicolon required -- at end of expression
mlr: cannot parse DSL expression.
Parse error on token "$y" at line 6 column 3.
Please check for missing semicolon.
Expected one of:
  ␚ ; > >> | ? || ^^ && =~ !=~ == != <=> >= < <= ^ & << >>> + - .+ .- *
  / // % .* ./ .// . ?? ??? ** [ [[ [[[


Miller has elif, not else if or elsif.

If-statement variable scoping

Miller is simple-minded about scoping local variables to blocks. If you have

  if (something) {
    x = 1
  } else {
    x = 2

then there are two x variable, each confined only to their enclosing curly braces; there is no hoisting out of the if and else blocks.

A suggestion is

  var x
  if (something) {
    x = 1
  } else {
    x = 2

Required curly braces

Bodies for all compound statements must be enclosed in curly braces, even if the body is a single statement:

mlr ... put 'if ($x == 1) $y = 2' # Syntax error
mlr ... put 'if ($x == 1) { $y = 2 }' # This is OK

No autoconvert to boolean

I.e. ints/strings/etc are neither "truthy" nor "falsy". Boolean tests in if/while/for/etc must always take a boolean expression: if (1) {...} results in the parse error Miller: conditional expression did not evaluate to boolean., Likewise if (x) {...}, unless x is a variable of boolean type. Please use if (x != 0) {...}, etc.

Integer-preserving arithmetic

As discussed on the arithmetic page the sum, difference, and product of two integers is again an integer, unless overflow occurs -- in which case Miller tries to convert to float in the least obtrusive way possible.

Likewise, while quotient and remainder are generally pythonic, the quotient and exponentiation of two integers is an integer when possible.

$ mlr repl -q
[mlr] 6/2

[mlr] typeof(6/2)

[mlr] 6/5

[mlr] typeof(6/5)

[mlr] typeof(7**8)

[mlr] typeof(7**80)

As seen in the previous example, print with multiple comma-delimited arguments fills in intervening spaces for you. If you want to avoid this, use the dot operator for string-concatenation instead.

mlr -n put -q '
  end {
    print "[", "a", "b", "c", "]";
    print "[" . "a" . "b" . "c" . "]";
[ a b c ]

Similarly, a final newline is printed for you; use printn to avoid this.

String literals with double quotes only

In some languages, like Ruby and Bash, string literals can be in single quotes or double quotes, where single quotes suppress \n converting to a newline character and double quotes allowing it: 'a\nb' prints as the four characters a, \, n, and b on one line; "a\nb" prints as an a on one line and a b on another.

In others, like Python and JavaScript, string literals can be in single quotes or double quotes, interchangeably -- so you can have "don't" or 'the "right" thing' as you wish.

In yet others, such as C/C++ and Java, string literals are in double auotes, like "abc", while single quotes are for character literals like 'a' or '\n'. In these, if s is a non-empty string, then s[0] is its first character.

In the Miller programming language:

  • String literals are always in double quotes, like "abc".
  • String-indexing/slicing always results in strings (even of length 1): "abc"[1:1] is the string "a", and there is no notion in the Miller programming language of a character type.
  • The single-quote character plays no role whatsoever in the grammar of the Miller programming language.
  • Single quotes are reserved for wrapping expressions at the system command line. For example, in mlr put '$message = "hello"' ..., the put verb gets the string $message = "hello"; the shell has consumed the outer single quotes by the time the Miller parser receives it.
  • Things are a little different on Windows, where """ sequences are sometimes necessary: see the Miller on Windows page.


Miller has a somewhat novel flavor of null data called absent: if a record has a field x then $y=$x creates a field y, but if it doesn't then the assignment is skipped. See the null-data page for more information.


See the maps page.

Arrays, including 1-up array indices

Arrays and strings are indexed starting with 1, not 0. This is discussed in detail on the arrays page and the strings page.

mlr --csv --from data/short.csv cat
mlr --csv --from data/short.csv put -q '
  @records[NR] = $*;
  end {
    for (i = 1; i <= NR; i += 1) {
      print "Record", i, "has word", @records[i]["word"];
Record 1 has word apple
Record 2 has word ball
Record 3 has word cat

Also, slices for arrays and strings are doubly inclusive: x[3:5] gets you elements 3, 4, and 5 of the array or string named x.

See the arrays page for more about arrays; see the strings page for more about strings.

Two-variable for-loops

Miller has a key-value loop flavor: whether x is a map or array, in for (k,v in x) { ... } the k will be bound to successive map keys (for maps) or 1-up array indices (for arrays), and the v will be bound to successive map values.

Semantics for one-variable for-loops

Miller also has a single-variable loop flavor. If x is a map then for (e in x) { ... } binds e to successive map keys (not values as in PHP). But if x is an array then for e in x) { ... } binds e to successive array values (not indices).

JSON parse, stringify, decode, and encode

Miller has the verbs json-parse and json-stringify, and the DSL functions json_parse and json_stringify. In some other languages these are called json_decode and json_encode.