SF3PRETTY 1L "29 April 1994" "University of Utah "

Table of contents

• NAME
• SYNOPSIS
• DESCRIPTION
• LANGUAGE RECOGNIZED
• OPTIONS
  • -a
  • -b
  • -c
  • -d
  • -h
  • -i
  • -k
  • -l
  • -m
  • -s
  • -u
• FILES
• SEE ALSO
• BUGS
• AUTHOR

NAME

SYNOPSIS

DESCRIPTION

In particular, it can be applied to the output of pretty(1L) to supply indentation for Fortran 77 block-IF statements, which pretty(1L) presently does not recognize as candidates for indentation. It is at least 20 times faster than pretty (1L), and processes about 275 lines/sec on a Sun 3/50 (15MHz Motorola 68020) system, so it can be routinely applied to code without serious time penalties.

Following the conventions of DEC Fortran compilers, an initial tab on a line blank-fills through column 6 if the next character is not a digit, and otherwise fills through column 5, placing the following digit in column 6 as a continuation character. Tabs embedded in lines are expanded by blank padding through the next column which is a multiple of 8 (thus preserving layout in comments). The output contains no tab characters.

In order to preserve neat indentation according to control structure level, long character strings are split, preferably at a blank. In executable statements, the concatenation operator, //, is used to merge them back into one string (usually at compile time). In FORMAT statements, comma separators are inserted. In DATA statements, where string expressions are not permitted by Fortran 77, the break is made at column 72, and the same string continues in column 6 of the next line. Because some compilers (erroneously) fail to pad input statements to column 72 before merging continuation lines, trailing blanks on a line in such a split string can be lost. To prevent this, if column 72 is blank when a string is wrapped, a hyphen will be output in column 73; it will be ignored by the compiler, but will protect against erroneous string truncation. Otherwise, trailing blanks are trimmed from the output.

LANGUAGE RECOGNIZED

To increase its utility, the analyzer also recognizes numerous language extensions present in commonly-used Fortran compilers. These are summarized in the following table. Note that some extensions are supported by multiple vendors.

Vendor

CDC

<letter>=, Rw, and nVw.d FORMAT items, nRxxxx right-adjusted Hollerith constant, "xxx" quoted strings

Cray

n/ FORMAT item

DEC

ACCEPT, DO ... ENDDO, IMPLICIT NONE, INCLUDE, REREAD, and TYPE statements, E, F, G, I, Ow.d, Q, Zw.d, and $ FORMAT items, long variable names, $ and % in variable names, tabbed source format, lower-case source format

IBM

&nnn alternate return specifier, NAMELIST I/O

UNIX

B, P, nR, SU, T, and nT FORMAT items

OPTIONS

-a

Display author and copyright information on stderr.

-bx

Set the comment character to x, which must be `*', `C', or `c'. If not given, the original comment characters are preserved.

-cx

Set the statement continuation character (for column 6) to x, which may be any character in the Fortran character set (upper-case letters, digits, space, = + - * ( ) / , . $ ' : ), except space or 0. The default is -cX.

-d#

Set optional space-deletion bits (see -s# below). A value -d0 removes internal defaults, so that input space will be exactly preserved in the output, except at the beginning of lines, which are indented according to the setting of the -i# switch. The default is -d0xffffffff, which will remove all whitespace around non-alphanumeric tokens (the default -s# switch setting restores appropriate spacing).

-h

Convert Fortran 66 Hollerith strings (like 6HO'hara) to Fortran 77 quoted character strings (like 'O''hara').

-i#

Set the statement indentation value to #. Values in the range 2..5 are suggested. The default is -i5 if this switch is not specified.

-k#,#,...

Set the declaration columns to the indicated values; the defaults are equivalent to -k26,39,52,65 matching the output of the Extended PFORT Verifier (pfort(1L)) declaration file. Commas, colons, semicolons, or whitespace may separate the numbers. If whitespace separators are used, the argument must be quoted to prevent it from being interpreted as multiple arguments.

If a negative number appears in the list, then all columns from the preceding number to the absolute value of the negative one are declaration columns. The shorthand of, say, -k26,-72 says that any column in the range 26 through 72 is a declaration column; in other words, variables will appear successfully without intervening space.

-lkeystring

Output in lower case the tokens of the type specified by the keystring value. The characters in the keystring may be in either letter case, and may occur in any order (even repeated). They may be one or more of the following:

C

constants (except quoted and Hollerith strings)

D

SFTRAN3 directives

F

FORMAT fields

I

Fortran INCLUDE statements

K

keywords

O

operators

V

variables, subroutines, functions, and other user-defined names

Letter case of comments and character strings is never altered. By default, input letter case is preserved. However, -ucdfikov could be used to convert all but comments and strings to upper case (ANSI standard), and -lCDFKIOV would correspondingly convert to lower case. If both switches are given with the same letter options, the last one specified is used.

If no keystring is specified, CDFKOV is assumed; this selects all options except the one affecting Fortran INCLUDE statements. Because they contain file names whose letter case may be significant on some operating systems, they are never modified unless explicitly requested.

-m

Display dynamic memory usage on stderr on termination.

-s#

The -d# and -s# switches set a group of bit flags for selecting spacing options. The -s switch selects space insertion, and the -d switch selects space deletion. Spacing is left unchanged for those tokens for which no option is given. The deletion values apply only to whitespace tokens present in the input stream; they do not remove space inserted by the spacing values.

The value is selected by or'ing together any of the following values, or by specifying the desired values in separate options, and it may be specified in decimal, octal (0nn), or hexadecimal (0xnnn):

0x00001

=

0x00002

+ -

0x00004

* /

0x00008

logical operators (.and., .eqv., .neqv., .not., .or., .xor.)

0x00010

relational operators (.eq., .ge., .gt., .le., .lt., .ne.)

0x00020

logical constants (.false., .true.)

0x00040

comma

0x00080

SFTRAN3 structure labels

0x00100

SFTRAN3 inline comments

0x00200

concatenation operator (//)

0x00400

exponentiation operator (**)

0x00800

colon

0x01000

ampersand

0x02000

left parenthesis

0x04000

right parenthesis

0x08000

semicolon

0x10000

keywords

0x20000

user-defined names

If a value 0 is specified, the internal defaults are cleared, so that original spacing will be preserved. The default value is -s0x183fb.

-ukeystring

Specify token types to be converted to upper case (see -l above).

To support consistent prettyprinting style, it is convenient to be able to supply standardly-used options in a default startup file. This file is called .sf3pretty on UNIX systems, or sf3pretty.opt on others, and it is searched for in the current directory, or if not found there, in the login directory. If the file is found, up to 72 characters of its first line are read and parsed into command line options. These should not be protected by shell quotes; for example, the command line argument '-c*' should appear simply as -c*.

FILES

Standard input and standard output.


./.sf3pretty

SEE ALSO

BUGS

AUTHOR

Nelson H. F. Beebe
Center for Scientific Computing
Department of Mathematics
University of Utah
Salt Lake City, UT 84112
USA

Tel: +1 801 581 5254
FAX: +1 801 581 4148

Email: beebe@math.utah.edu