13. `FORMAT` & edit descriptors

A format describes how data is laid out as text — field widths, decimal places, spacing, literal text. It pairs with a formatted READ/WRITE (Chapter 12) and is the most detailed part of the language (§13).

Where a format comes from

Three ways to supply one:

      WRITE (6, 10) N, X
   10 FORMAT (I5, F8.3)               ! a labelled FORMAT statement

      WRITE (6, '(I5, F8.3)') N, X    ! a character-constant format, inline

      CHARACTER*20 FM
      FM = '(I5, F8.3)'
      WRITE (6, FM) N, X              ! a character variable holding a format

A format is a parenthesized list of edit descriptors separated by commas.

Numeric edit descriptors (repeatable)

Descriptor	For	Example	Output of the value
`Iw`	integer in width `w`	`I5` of `42`	`␣␣␣42`
`Iw.m`	integer, ≥ `m` digits (zero-filled)	`I5.3` of `42`	`␣␣042`
`Fw.d`	fixed-point real, `d` decimals	`F8.3` of `3.14159`	`␣␣␣3.142`
`Ew.d`	scientific real	`E12.4` of `1.5`	`␣␣0.1500E+01`
`Ew.dEe`	scientific, exponent width `e`	`E12.4E3`	`…E+001`
`Dw.d`	double-precision scientific (like `E`)	`D12.4`	`…D+01`
`Gw.d`	general — picks `F` or `E` by magnitude	`G12.4`	(F- or E-form)

(␣ marks a blank.) Output is right-justified in the field; if the value doesn't fit, the field fills with asterisks (****). On input, leading blanks are ignored, a + is optional, and an all-blank field reads as zero.

      WRITE (6, '(I5, F8.3)') 42, 3.14159     ! ->  '   42   3.142'
      WRITE (6, '(E12.4)')    1.5             ! ->  '  0.1500E+01'

Logical and character (repeatable)

Descriptor	For	Notes
`Lw`	logical	outputs `w−1` blanks then `T` or `F`
`A`	character	uses the item's own declared length
`Aw`	character in width `w`	right-justify if `w` > len; leftmost `w` chars if `w` ≤ len

      CHARACTER*3 C
      C = 'HI'
      WRITE (6, '(A)') C            ! -> 'HI '   (the item's length, 3)

Literal text and Hollerith (output)

Descriptor	Effect
`'text'`	print the quoted text literally
`nHtext`	print the `n` characters after `H` (the old Hollerith form)

      WRITE (6, '('' answer ='', I4)') 42      ! ->  ' answer =  42'

Positional & control (non-repeatable)

Descriptor	Effect
`nX`	skip `n` positions (blanks on output)
`Tc`	move to absolute column `c`
`TLc` / `TRc`	move left / right `c` columns
`/`	end this record, start the next (a line break)
`:`	stop format processing if the I/O list is exhausted
`S` `SP` `SS`	sign control: default / always print `+` / suppress `+`
`kP`	scale factor: shift the decimal point by `k`
`BN` `BZ`	on input, treat blanks in a numeric field as null / as zeros

      WRITE (6, '(I3, /, I3)') 1, 2      ! two lines: ' 1' then ' 2'

Repeat counts and groups

A descriptor may be repeated with a leading count, and a parenthesized group may be repeated as a unit:

      FORMAT (3I5)              ! same as I5, I5, I5
      FORMAT (2(I3, F6.2))      ! the group (I3,F6.2) twice

Format reversion

If the I/O list has more items than the format has descriptors, the format reverts: a new record starts and processing resumes at the last open parenthesis (the last group) (§13.3). This is how one FORMAT prints an arbitrarily long list, a record per group:

      WRITE (6, '(5I4)') (J, J = 1, 12)   ! 5 per line, reverting -> 3 lines

Carriage control (printing)

When a formatted record is sent to a device that prints, the first character of each record is not printed — it controls vertical spacing (§12.9.5.2.3):

First char	Effect
blank	advance one line (normal)
`0`	advance two lines (blank line before)
`1`	advance to the top of a new page
`+`	no advance (overprint the previous line)

This is the famous gotcha: on a printing device, WRITE (6,'(I3)') 100 prints 00, because the 1 is eaten as carriage control (and starts a new page!). The cure is to begin output formats with a blank — 1X or ' ':

      WRITE (6, '('' '', I3)') 100      ! leading blank = single-space; prints ' 100'

Whether a given unit prints (interprets that first character) is left to the processor — see the forterp notes.

List-directed formatting

FMT = * skips descriptors entirely and formats each value sensibly, separated by blanks (output) or by blanks/commas/slashes (input) — see Chapter 12.

forterp notes. The full edit-descriptor family is implemented (quick reference: Appendix B), including Iw.m, the scale factor, S/SP/SS, BN/BZ, the exact-exponent Ew.dEe, and reversion. The carriage-control question — does the first character get consumed? — is answered per dialect, because the standard leaves "which devices print" to the processor: under F77 standard output is a terminal (the first character is ordinary data, matching gfortran), while under F66/FORTRAN10 it is a line printer (first character consumed as carriage control). The carriage_control engine flag overrides the dialect default. See Appendix D.