NAME

rw — blockwise input/output

SYNOPSIS

rw	[-afhPstvx] [-b block-size] [-c count] [-I input-offset] [-i input-file] [-O output-offset] [-o output-file] [-p period] [-r input-block-size] [-w output-block-size]

DESCRIPTION

rw reads blocks from the standard input and copies them to the standard output until the end of the standard input. The input and output block sizes default to appropriate values for efficiently reading from the input and writing to the output. Input and output will be done as aligned as possible on the block size boundaries. The final input block can be partial. Output blocks are written whenever enough input blocks have been read, or partially written whenever the end of the input is reached. The output file is not truncated after the copy is done. Byte quantities can be specified as a non-negative count of bytes in decimal format, hexadecimal format (with leading 0x), or octal format (with leading 0); optionally followed by any amount of whitespace, and then suffixed with any of the following case-insensitive suffixes that multiplies the specified quantity by that magnitude:

B

Magnitude of a byte (1 byte).

K, KiB

Magnitude of kibibytes (1,024 bytes).

M, MiB

Magnitude of mebibytes (1,048,576 bytes).

G, GiB

Magnitude of gibibytes (1,073,741,824 bytes).

T, TiB

Magnitude of tebibytes (1,099,511,627,776 bytes).

P, PiB

Magnitude of pebibytes (1,125,899,906,842,624 bytes).

E, EiB

Magnitude of exbibytes (1,152,921,504,606,846,976 bytes).

r

Magnitude of input blocks (the default input block size when setting block sizes).

w

Magnitude of output blocks (the default output block size when setting block sizes

x

Magnitude of input and output blocks (if they have the same size) (the default block size when setting block sizes).

The options are as follows:

-a

In combination with -o, open the output file in append mode. This option must be used with -o and is incompatible with -t. The output offset is set to the size of the output file.

-b block-size

Set both the input and output block sizes to the byte quantity specified by block-size.

-c count

Stop after copying the byte quantity specified by count, in addition to stopping when the end of the input is reached. If the count starts with a leading ‘-’, stop that many bytes before the end of the input (only works if the input size is known).

-f

Continue as much as possible in the event of I/O errors and exit unsuccessfully afterwards. For each input failure, skip forward to the next input block or the next native-size input block or the end of the file (whichever comes first), write an error to the standard error, and replace the failed input with NUL bytes when writing it to the output. For each output failure, skip forward to the next output block or the next native-size output block (whichever comes first) and write an error to the standard error. The native-size input and output blocks are those defined by the preferred input/output size for the input and output. This option only works for the input if it is seekable and the input size is known. This option only works for the output if it is seekable (and -a is not set). Beware that the default preferred input/output sizes may be larger than the underlying storage sector sizes: If this option is used, the -r and -w options should be set to the appropriate input/output sector sizes, or more than just the bad sector may be skipped.

-h

Write statistics in the human readable format where byte amounts and time amounts are formatted according to their magnitude as described in the DIAGNOSTICS section.

-I offset

Skip past the first offset bytes in the input before the copying begins. If the offset starts with a leading ‘-’ it is interpreted as that many bytes before the end of the input (if the size is known), and if it starts with a leading ‘+’ it is interpreted as that many bytes after the end of the input (if the size is known). If the input is not seekable, the first offset bytes are read and discarded before the copying begins. If the offset is not a multiple of the input block size, the first input block is reduced in size such that it ends at an input-block-size-aligned position in the input.

-i input-file

Read the input from input-file instead of the standard input.

-O offset

Seek past the first offset bytes in the output before the copying begins. If the offset starts with a leading ‘-’ it is interpreted as that many bytes before the end of the output (if the size is known), and if it starts with a leading ‘+’ it is interpreted as that many bytes after the end of the output (if the size is known). If the output is not seekable, the number of NUL bytes specified in offset are written to the output before the copying begins. This option cannot be set to a non-zero value if -a is set. If the offset is not a multiple of the output block size, the first output block is reduced in size such that it ends at a output block size aligned position in the output.

-o output-file

Write the output to output-file instead of the standard output, creating the file if it doesn't exist. If output-file already exists, the existing data is not discarded. Use -t if you want to truncate the output afterwards.

-P

Pad the final output block with NUL bytes, such that the final output offset (counting the initial offset with -O) is a multiple of the output block size.

-p period

Write occasional statistics to the standard error during the transfer and on completion, or when being terminated by SIGINT. period specifies the period in whole seconds at which statistics are written out. Statistics are written for every read and write if the period is zero. The format is described in the DIAGNOSTICS section.

-r input-block-size

Set the input block size to the byte quantity specified by input-block-size.

-s

Sync the output on successful completion.

-t

Truncate the output to the final output position after the copy has completed. This option requires the output to be truncatable. This option is incompatible with -a.

-v

Write statistics to the standard error upon completion, or when being terminated by SIGINT. The format is described in the DIAGNOSTICS section.

-w output-block-size

Set the output block size to the byte quantity specified by output-block-size.

-x

In combination with -o, fail if the output file already exists.

ASYNCHRONOUS EVENTS

SIGINT

If -v or -p is set, abort the copy, write statistics to the standard error, and then exit as if killed by SIGINT.

SIGUSR1

Write statistics to the standard error and continue the copy. If SIGUSR1 is not ignored, this handler is installed and this signal is unblocked. To use this signal without a race condition before the signal handler is installed (as SIGUSR1 is deadly by default), block the signal before loading this program. To disable the handling of this signal, ignore the signal before loading this program.

EXIT STATUS

rw will exit 0 on success and non-zero otherwise.

EXAMPLES

Copy from the standard input to the standard output:

rw

Copy the first 256 bytes from the input to the output:

rw -c 256

Copy from the input file foo to the beginning of the output file bar (preserving any data in the output file beyond the final output position after the copy is finished).

rw -i foo -o bar

Copy from the input file to the beginning of the output file, truncating the output file to the final output position afterwards:

rw -i foo -o bar -t

Copy from the input file foo to the beginning of the output block device /dev/bar (preserving any existing data on the output block device beyond the copied area), while writing progress statistics every 10 seconds in the human readable format, and sync the output block device afterwards:

rw -i foo -o /dev/bar -p 10 -h -s

Skip the first 512 bytes of the input, and then append the next 1024 bytes to the output file bar:

rw -I 512 -c 1024 -o bar -a

Copy 2 KiB from offset 768 in the input file foo to offset 256 MiB in the output file bar.

rw -c 2K -i foo -I 768 -o bar -O 256M

Copy from sector 32 and 4 sectors onwards from a block device /dev/foo (with the sector size being 512 bytes) to the output file bar:

rw -r 512 -i /dev/foo -I 32r -c 4r -o bar

With a block size of 4096 bytes, copy 64 blocks from the input from offset 32 blocks in the input to offset 65536 blocks in the output:

rw -b 4096 -c 64x -I 32x -O 65536x

Back up the /dev/foo block device (with the sector size being 512 bytes) to the bar output file, continuing despite I/O errors by writing error messages to the standard error and writing NUL bytes to the output instead, truncating the output file to the size of the input, writing progress statistics every 10 seconds in the human readable format to the standard error:

rw -f -i /dev/foo -r 512 -o bar -t -p 10 -h

With the input block size of 512 bytes and the output block size of 8192 bytes, copy 16384 input blocks from input block 65536 onwards to output block 1048576:

rw -r 512 -w 8192 -c 16384r -I 65536r -O 1048576w

Copy 512 bytes from 1024 bytes before the end of the input to 2048 bytes after the current size of the output file:

rw -c 512 -I -1024 -o bar -O +2048

Skip the first 100 bytes of the input and copy until 200 bytes are left in the input file:

rw -i foo -I 100 -c -200

DIAGNOSTICS

Statistics about the copy are written to the standard error upon completion if either -v or -p are set; occasionally if -p is set; upon SIGINT (if not ignored when the program was loaded) if -v is set; and upon SIGUSR1 (if not ignored when the program was loaded). The statistics are in this format:

<time-elapsed> s <done> B / <total> B <percent>% <speed> B/s <time-left> s

time-elapsed is the number of seconds since the copying began. done is the number of bytes copied so far. total is an estimate of how many bytes will be copied, or ‘?’ if not known. percent is how many percent complete the copy is, or ‘?’ if not known. speed is the average speed of copying so far in bytes per second, or ‘?’ if it is too early to tell. time-left is the number of seconds left, assuming the remaining data is copied at the current average speed, or ‘?’ is not known. For instance, the statistics could look like this:

7 s 714682368 B / 1238364160 B 57% 102097481 B/s 5 s

The statistics are printed with human readable byte units (B, KiB, MiB, GiB, TiB, PiB, EiB) and time units (s, m, h, d) if the -h option is set:

7 s 714.4 MiB / 1.1 GiB 60% 102.0 MiB/s 4 s

SEE ALSO

cat(1), cp(1), dd(1)

HISTORY

rw originally appeared in Sortix 1.1. rw is similar to dd(1), but has a distinct design and improvements:

• The command line options use the conventional option format.
• The output file is not truncated by default. One has to use -t.
• The input and output block sizes default to the preferred I/O block sizes instead of 512 bytes.
  The -c, -I, and -O options accept byte quantities by default instead of block counts, but can be specified in block counts by using the ‘r, w’, and ‘x’ suffixes.
• Statistics are not written by default. One has to use -v or -p. The statistics contain more useful information and is machine readable as it contains no localized information. A human readable statistics format is available using -h. Statistics can occasionally be written out using -p.
• There is no support for converting ASCII to EBCDIC, converting ASCII to a different EBCDIC, EBCDIC to ASCII, swapping pairs of bytes, converting the bytes to lower-case or upper-case, converting line-delimited data into fixed-size blocks, or converting fixed-sized blocks into line-delimited data.
• Offsets can be specified relative to the end of the input/output.
• Input errors stop the copying immediately rather than writing out a partial output block.