Generalized I/O

Mystery of stars• I/O statements took the form

read *, x, y,z print *, a,b,c

• Such I/O is said to be List-directed or in free format• user has no control over the appearance of I/O• No flexibility over the I/O devices

• Keyboard input• Display unit output

• * indicates no control or default format

Mystery resolved

• * stands for (*,*)

• In general, it will be (arg1,arg2)

• arg1 used to select the I/O device– * indicates default devices (keyboard and

monitor)

• arg2 controls the I/O formatting– It is a string constant– * means default formatting

Files• I/O through files is the most convenient• Standard I/O is very painful for large data• Large amount of data can be stored and

retrieved with ease• Output of one program can be used as input

to another• The programs need not be running at the

same time• Once created they remain - permanent

memory• Files are high level software abstractions of

these permanent memory devices• They are called as secondary memory

Secondary Memory• permanent, slow and large volume of data• Some kind of magnetic memory –• no power required to retain unlike main

memory• different kinds - Hard Disks, Floppy Disc,

CDs, Magnetic Tapes (old)• Random or Sequential access• details of devices abstracted from application

programmers• Files are the abstractions• Std. I/O are also files

File Organization• Fortran programmer's view

• A simple file is a linear sequence of records

• Each record is a meaningful piece of data

• Files are read or written one record at a time

• Each record is typically a line of data–creation of file determines type of data

• Files can be sequential or random access

• Random access files require keys

File Operations• Opening a file

• writing or reading a record

• Rewinding a file

• Closing a file

• Valid sequence– (open (read/write/rewind)* close)*

• File operations translated to OS calls

• They can fail and graceful recovery possible

File Opening• Simplest example is:

open(unit = 23, file = "input.dat")• This commands associates the file input.dat with the

number 23• This number can be used for reading and writing onto

the file

• Example: read(23, *) Nwrite(23, *) M

• Unit numbers are system dependent (usually from 1 to 100)

• unit and file are named parameters• Contrast this with positional parameters

Open attributes• Many file attributes assume default values in the

above• to override these, additional parameters to be set• action = READ, WRITE or READWRITE

(default)• status = OLD, NEW, REPLACE (default),

SCRATCH, UNKNOWN• ios = x, x is assigned a value at the end of the call

0 - if the operation is successful

> 0 - otherwise• access = "sequential" (default) or "direct"

Some more attributes

• form = "formatted" (default) or

"unformatted“– in unformatted form, data stored in some internal

representation– can not be edited

• recl = length of each record– required for direct access files

• position = "rewind", "append"

Exampleinteger:: ierr

open(unit=23, file = "in.dat", status = "old",

action = "Read“, iostat = ierr)

– opening the existing file "in.dat", for read– ierr gets a value > 0 iff failure in opening

Another Exampleinteger:: ierrinteger:: ut = 34open(unit= ut, file = "out.dat", status = "new", action = "write“, iostat = ierr)

– creating a new file "out.dat– Status = replace would replace the file

One more Example

open(unit= 15, status = "scratch", iostat = ierr)

– no file name specified - error otherwise– default action - reading and writing temporary data– erased after the program terminates

Closing a file• Close(close_list)• Close_list

– must specify the unit number– may specify other clauses

• the number is released and can be reused in other open statements

• repeated opening and closing of non scratch files possible

• scratch file contents lost once they are closed

Reading and Writing• files contain sequence of records• Each record is read/written using a single read or

write operation• So we can say that each record occupies a single line• Every read statement refers to the successive

records• Reference to the file through the unit number• example:

read(unit = 16, *) x, y, z

write(unit = 25, *) "test output=", a, b, c• read and write one record to the appropriate file

Read Pointer• There can be errors while reading a file

• There is a read pointer that indicates the next record to be read

• Every read moves the read pointer to the next record (in sequential access)

Read Pointer

Read Error• The read pointer may be at the end

• use istat clause for detecting end of file (EOF)

Read(16,*, iostat = err) x, y, z

• If err is non zero then there is an error

• Use always iostat clause

• otherwise, program will abort

Other parameters• For direct access files, use the argument

rec = n• This reads/writes the nth record

Example:

write(23,*,iostat = err, rec = 15) i,j

– write values of i, j to the record 15 of the associated file

Program bubble_sort

implicit none

integer:: N, i, rt_end, index, temp! N - the total number of items in the array

! i, j - array index variables integer, dimension(100):: A integer:: err,errf, werr

open(unit=1, file = "in.dat", & status = "old", iostat = err) if (err/=0) then print *, "file opening error" else

outer: do

Program Continued

read (unit =1, iostat = errf, fmt = * ) Nread (1, *, iostat = errf) (A(i),i=1,N)

if (errf /= 0) exit rt_end = N inner: do if (rt_end <= 1) exit index = 0 do i = 1, (rt_end-1) if (A(i) > A(i+1)) then temp = A(i) A(i) = A(i+1) A(i+1) = temp index = i endif enddo rt_end = index

end do inner

Program Continued

open(unit=2, file = "out.dat", & status = "old", access = "sequential", &

iostat = werr)

print *, (A(i),i=1,N)

write (2, fmt = *) (A(i),i=1,N)

end do outer

endif

close(unit=1)

end program

I/O Formatting• So far we have used List-Directed I/O Statements

(Free format)

print *, A,B,C

read *, A,B• Fine for small programs and small number of inputs

and outputs• Many data processing applications manipulate large

collection of inputs and outputs– Tabulating marks and results– preparing mark sheets and reservation charts

• Pretty printing and Formatted input very essential• Fortran has elaborate features for formatting I/O

Format Strings• I/O formats can be specified by using format strings

print fs, A, B,C

read fs, C,D• fs identifies a format string

– a constant string– a string variable– a label that labels a format statement containing the

format string

• Here is an example format strings

print "I4, I4, I4", A, B, C

• This prints A, B and C in a row with each being given 4 decimal places

Edit Descriptor• I4 is an edit descriptor• Format string contains one or more edit descriptors• Very many edit descriptors that help print different

types of variables• Alternate format strings

Character(len=20):: strstr = "(I4, I4, I4, I4)”write(*, str) a,b,c,d

read *, 100, a, b, c, d100 format ("I4,I4,I4,I4")

print "4I4", a, b,c,d

I/O buffer

• I/O done through buffers

• Buffer is a temporary storage space where every line of characters is composed before doing I/O

• Line-oriented I/O

• For reading – input process builds the buffer and gives the buffer content to the program

• Every print statement composes the output in the output buffer completely before sending for printing on the device

• Buffers are read/written as per the format statement

Output Buffer• Usually 133 characters long (line-printer output)• First character is a control character

– 1 skip to next page– Blank single spacing (default when * is used)– 0 double spacing– + no space – overwriting

• The rest of the characters are printed• Make sure that first character is the intended control

character print (*,100) n 100 format (I3)Behaviour of print statement when n = 25 and n =125

Some descriptors• Iw - w columns reserved for printing an

integer value includes - if the number is negative

• 5Iw - 5 successive w columns reserved for printing/reading 5 values

• Iw.m - minimum m digits need to be presented

• If w not enough then * printed• If the number is less than m digits -

leading zeros added

Format Descriptors• rFw.d

r - repeat count

w - number of columns for the entire number

(including decimal point and -ve sign)

d - number of columns to the right of the decimal point

• Number appears right-justified

• Trailing zeros if the no. of digits after decimal point is less than d

• rounded off if d is more

• if (w-d-1) is less, then asterisks printed

E-descriptor

rEw.d

Printing is done in E notation

Print "E11.4", A

will print -0.1234E-02 when the value of A is

1.234 * 10 ** -1

Make sure that

w > = d + 7

A-descriptorrA or rAw

if w is more then A is right justifiesif w is less than only first w characters of A are printed

X descriptor

nX inserts n spaces in the output

T descriptor

Tc moves to the column number c

rLw

used for printing logical variables - w columns are used output will be either T or F, right justified

RepetitionRepetition can be applied to groups of descriptors

2(I6, 2X, 2F10.2)

is used to print/read two groups, each consisting of an integer (6 digits)and two floating point numbers of size 10

/ operator used for breaking the line

A single write statement can print on more than one line

Descriptor Association• print fs, var_list• association of format descriptor to a variable is done

from left to right• if for every descriptor (except print controls) there is

a variable and vice versa, then association is simple• if less variables then rest of the descriptors

discarded• if more variables then the printing starts from a fresh

line for rest of the variables• format strings for these variables start from right

most open parenthesis that does not have repetition count

Example print (8, 60) j,k,l,m,n

60 format (1X,"The value” = , /, (1X, ‘New Line', 2(3X, I5)))

The output is

1 16 24 (column number not printed)

The value =

New Line 1 2

New Line 3 4

New Line 5

where j - n has the value 1 - 5

Other aspects

• Formatted reading is similar (see the books/manuals for details)

• Format strings can be specified in generalized file-based I/O

• Include the format string in the argument fmt