Labels

A label is a name for the figure on a line: the Rent in Rent: $1200. It is what turns a column of numbers into a budget you can read, and the engine keeps it as text and works out the rest of the line. The answer is the figure alone, so a total above or a lone sum under the column adds up the amounts, never the words.

A label before a colon

Words and then a colon make a label, whatever follows the colon: an amount, or a whole expression.

Rent: $1200 // $1,200.00
pi approximation: 355/113 // 3.14
total: 5 + 3 // 8
Groceries: 45 // 45

The expression after the colon is read exactly as it would be on a line of its own, colons of its own included. A range inside a call, the 1000:1002 of total(1000:1002) (the whole numbers from 1,000 to 1,002), keeps its meaning after a label: a label stands at the start of the line, never inside a bracket, so a colon inside one is never taken for the label’s. That line used to be refused as the time of day “1000:1002”.

Total: total(1000:1002) // 3,003
Total: total(1,000:1,002) // 3,003
Cost: total(10:12) // 33
Squares: sum(x^2, 1:3) // 14

One letter is a word too, so x: 3 is a label, the way A: 40 and B: 55 label a ledger. The label is only a name for the figure: it never reads or changes a variable called the same, so with x = 2 above it, x: 3 is 3 and x is still 2. A colon has no other meaning here to fall back on: a ratio is written ratio(x, 3) (see ratios), a colon between two numbers is a clock time, and a variable is set with :x = 3 or x = 3.

x = 2 // 2
x:3 // 3
x + 1 // 3

What can stand before the colon

A label is a name, the way a ledger names a figure: words, with the numbers and joining marks a name has. Week 12, Year-end, Cost/unit and Food + drink are all names, so each is a label and the figure after the colon is the answer:

Week 12: 75 // 75
Item 2: 45 // 45
Weeks 1-2: 40 // 40
Year-end: 5 // 5
Food + drink: $40 // $40.00
Done?: 5 // 5
Orders over $100: 12 // 12

A number in a name and the figure after it can look like a clock time: in Item 2: 45 the 2 and the 45 are also a quarter to three. What tells them apart is the space. A time is written with its colon touching the minutes, and a label’s colon has a space after it, as in prose, so when a name stands before the number and a space follows the colon, the colon is the label’s and the figure is the answer. Without the space, Item 2:45 is the word Item beside the time 2:45, which the line cannot work out, and with no name before the number, 9: 30 is still half past nine:

Room 4: 12 // 12
Item 2:45 // ERROR: Expected an operator or the end of the line, but found "2:45"
9: 30 // 9:30:00 AM

A name here is a word with only the numbers and joining marks (-, /) of a name between it and the colon, as in Weeks 1-2 or Part 1/2. A word that leads into a time is not a name, so before 9: 30 keeps its time, and nor is an operator or =: x = 5: 6 sets x to the time 5:06.

A colon has other jobs too, and when the text before it is doing something other than naming, reading it as a label would throw that text away and answer with whatever followed: 1 + 24:00 used to answer 0. So the text before the colon is refused, with what it is instead, when it is:

• Part of a time. A colon between two numbers belongs to a clock time, so a number that starts the line, follows an operator or follows a label’s colon is an operand, and a pair that is no real time (24:00, a seconds field of 99) is refused as one. A number that follows a word, as in Week 12, is part of the name.
• A figure in another script’s digits. Numbers are read in the digits 0 to 9 only. A figure written in another script’s digits (Arabic-Indic ٢٤, Devanagari २४, fullwidth １２) is not read as a number, and stood where a number would start, it is not a name either: ٢٤:00 is the time it looks like. It is refused by name, spelled in 0 to 9. After a word it is part of the name, as a number is (Week ٢: 5).
• A choice written with ? and :. There is no such operator; the refusal spells the line as the if ... then ... else the engine reads (see conditionals).
• A comparison. >, <, >=, <=, == or != written as a symbol make the text a condition, not a name. The same idea written in words (Orders over $100) is a name and stays a label.
• A figure with an invisible character in it. Some characters take up no space on screen: the direction marks and overrides that tell a display which way text runs (used for right-to-left scripts such as Arabic and Hebrew), a zero-width joiner, a soft hyphen. Typed against a number, one makes it a word the engine cannot read as a number, so stood where a number would start, it is not a name either: <U+202E>24:00 is the time 24:00 with a right-to-left override in front of it. It is refused by name, with the character written as its code point, so the reader can delete it. A direction mark or override gets the same refusal it gets in any name.
• A calculation with no word in it, such as (1+2), which names nothing. A bracketed figure such as (24) or [24] is one too: brackets with no word beside them are an expression, not a name.

1 + 24:00 // ERROR: "24:00" is not a valid time
1:23:99 // ERROR: "1:23:99" is not a valid time
Total: 24:00 // ERROR: "24:00" is not a valid time
Total: 1000:1002 // ERROR: "1000:1002" is not a valid time
true ? 25 : 30 // ERROR: There is no choice written with "?" and ":": write if true then 25 else 30
a > b: 1 // ERROR: "a > b" before the colon is a comparison, not a label: a label names the figure in words, and a choice is written if ... then ... else
(1+2): 5 // ERROR: "(1+2)" before the colon is a calculation, not a label: a label names the figure in words
(24):00 // ERROR: "(24)" before the colon is a calculation, not a label: a label names the figure in words
Total (2026): 500 // 500

A bracketed figure before a time’s colon used to be read as a label: (24):00 was the label (24) and answered the 0 after the colon, and [24]:00 and (9):30 did the same. A bracket beside a word is still part of a name, as in Total (2026) above, and (net): 5 is a label too.

A figure with an invisible character in it was read the same way, since the character made the figure a word: a right-to-left override in front of 24:00 answered 0. Each is now refused with the character named, as the zero-width joiner here:

‍24:00 // ERROR: "<U+200D>24" holds U+200D (zero width joiner), an invisible character, so it is read as a word and not as the number 24. A number cannot hold one: delete it and type the number again.

The boundary: a label of words keeps these characters, since a label is text, and a right-to-left label needs the direction marks to show correctly, so Rent followed by a right-to-left mark and : 5 is still 5. A figure after a word is part of the name, as it is for a number in 0 to 9. The zero-width space and the byte-order mark are read as spaces wherever they stand, so a time with one in front is the time it looks like.

The figure after a label is the start of the expression, so Total: 24:00 is the time 24:00 and is refused as 24:00 alone is. It used to be read as a second label, 24, and answered 0, and Total: 1000:1002 answered 1,002. A pair like 1000:1002 is no time, and at the start of an expression it is no range either (a range is read only as the list of sum, prod, map or reduce), so it is refused in the same words as on a line of its own; Total: sum(1000:1002) adds the range up.

A figure in another script’s digits used to be read as a label, since the engine reads such a figure as a word: ٢٤:00 answered the 00 after the colon, 0, and Total: ٢٤:00 did the same. The line is now refused by name, with the figure written in the digits the engine reads, so retyping it gives the answer:

٢٤:00 // ERROR: "٢٤" is written in digits the engine does not read: numbers are written in the digits 0 to 9, as in 24
Total: ٢٤:00 // ERROR: "٢٤" is written in digits the engine does not read: numbers are written in the digits 0 to 9, as in 24
Start: ٩:٣٠ // ERROR: "٩" is written in digits the engine does not read: numbers are written in the digits 0 to 9, as in 9
Week ٢: 5 // 5

The boundary: only a figure that stands before a colon, where a label would otherwise swallow it, is refused this way. The same figure elsewhere on a line is a word the engine does not know, and is refused as one (٢٤ + 1 names ٢٤ as an undefined name), so no line answers a number for digits the engine cannot read. Reading those digits as numbers is a larger change to how a line is read, and is not made here.

When the expression after a label cannot be worked out, the line says what is wrong with the expression, in the words that expression gets on a line of its own. It used to report only that the label’s colon was unexpected, which is true of the line but says nothing a reader can act on.

Total: average(10:12) // ERROR: In average(...), 10:12 is a clock time, not a range, and a time cannot be averaged: a colon between two numbers is a range only as the list of sum, prod, map or reduce. To average numbers, list them with commas, as in average(1, 2, 3).
Total: (1 + 2 // ERROR: The line ends where ")" was expected
Rent: $1200 + // ERROR: The line ends after "+", where a value was expected

A colon followed by = keeps the line’s own wording, since x := 5 is an assignment written another language’s way, and that refusal says to assign with = on its own.

The boundary: arithmetic between words stays a label (Food + drink, Year-end, Q1/Q2), since that is how ledgers name things and the figure after the colon is still the one the reader wrote. A question mark that ends the label (Done?: 5) is part of the name, but one with more words after it (Paid? yes: 5) reads as the start of a choice and is refused. And a pair after an operator is read as a time only when it is written as one, with the colon touching both numbers: in Score >= 90: 12 the colon has a space after it, so the text before it is judged as a label, and refused as a comparison rather than as the time 90:12.

A label without the colon

The colon is the part people leave out, and other notepads read a line without it. A run of words followed by one amount of money or one quantity reads the same way: the words are the label and the amount is the answer.

Rent $1200 // $1,200.00
Flight to Paris $450 // $450.00
Petrol 40 l // 40.00 l
Fees 45 EUR // €45.00

So a budget typed that way totals as it is written:

Rent $1200 // $1,200.00
Food $300 // $300.00
sum // $1,500.00

The label’s words can include a word the engine also reads, such as the to in Flight to Paris, since it is the amount at the end of the line that makes the line a label. A markdown list marker in front and a #tag or a // comment after are read as they are anywhere else.

Which lines stay prose

A note is mostly prose, and a sentence often has a number in it. Only a line that does not work out as it stands can become a label, and only in the shape a ledger line has, so these stay sentences:

I walked 5 km to the shop // ERROR: Expected an operator or the end of the line, but found "walked"
Meeting at 3 in room 4 // ERROR: Expected an operator or the end of the line, but found "at"
Chapter 12 // ERROR: Expected an operator or the end of the line, but found "12"
Back in 5 min // ERROR: Expected an operator or the end of the line, but found "5"
Remember the $5 // ERROR: Expected an operator or the end of the line, but found "the"

What makes the difference, line by line:

• The amount ends the line. Anything after it (5 km to the shop) leaves the line alone.
• The amount is money or a quantity. A bare number after words is not taken, because nothing tells Groceries 45 from Chapter 12, Room 4 or Page 3, which are names rather than amounts. Write the colon for a plain number: Groceries: 45.
• The label is only words and ends on one that names something. A number inside it, or a word that only leads into the amount (in, at, to, the, my, about, is), makes the line a sentence about the amount rather than a line naming it.

The boundary: a line that already works out is never read as a label, and nor is one that parses and fails as it runs. Refund -$50 is the subtraction Refund - $50, which fails because nothing defines Refund; write the colon for a negative amount.

Refund -$50 // ERROR: Undefined variable: Refund
Refund: -$50 // -$50.00

A line of words ending in an amount is read as a label wherever it stands, a sentence such as I paid $5 included, since that is the ledger’s own shape. The totals that read a column, and how a lone sum or total totals the block above it, are on line references.