Искусство программирования на языке сценариев командной оболочки

: <

The key (no pun intended) to a Unified Content File System (UCFS)

is to distinguish the files in the system based on their content.

Distinguishing files by their name is just, so, 20th Century.

The content is distinguished by computing a checksum of that content.

This version uses the md5sum program to generate a 128 bit checksum

representative of the file's contents.

There is a chance that two files having different content might

generate the same checksum using md5sum (or any checksum). Should

that become a problem, then the use of md5sum can be replace by a

cyrptographic signature. But until then...

The md5sum program is documented as outputting three fields (and it

does), but when read it appears as two fields (array elements). This

is caused by the lack of whitespace between the second and third field.

So this function gropes the md5sum output and returns:

[0] 32 character checksum in hexidecimal (UCFS filename)

[1] Single character: ' ' text file, '*' binary file

[2] Filesystem (20th Century Style) name

Note: That name may be the character '-' indicating STDIN read.

DigestFilesDoc

DigestFile()

{

local if=0 # Default, variable name

local -a T1 T2

case "$#" in

3) case "$1" in

-if) if=1 ; shift ;;

* ) return 1 ;;

esac ;;

2) : ;; # Poor man's "continue"

*) return 1 ;;

esac

case $if in

0) eval T1=\( \"\$\{$1\[@\]\}\" \)

T2=( $(echo ${T1[@]} | md5sum -) )

;;

1) T2=( $(md5sum $1) )

;;

esac

case ${#T2[@]} in

0) return 1 ;;

1) return 1 ;;

2) case ${T2[1]:0:1} in # SanScrit-2.0.5

\*) T2[${#T2[@]}]=${T2[1]:1}

T2[1]=\*

;;

*) T2[${#T2[@]}]=${T2[1]}

T2[1]=" "

;;

esac

;;

3) : ;; # Assume it worked

*) return 1 ;;

esac

local -i len=${#T2[0]}

if [ $len -ne 32 ] ; then return 1 ; fi

eval $2=\( \"\$\{T2\[@\]\}\" \)

}

# # # # # Locate File # # # # #

#

# LocateFile [-l] FileName Location-Array-Name

# or

# LocateFile [-l] -of FileName Location-Array-FileName

# # # # #

# A file location is Filesystem-id and inode-number

# Here document used as a comment block.

: <

Based on stat, version 2.2

stat -t and stat -lt fields

[0] name

[1] Total size

File - number of bytes

Symbolic link - string length of pathname

[2] Number of (512 byte) blocks allocated

[3] File type and Access rights (hex)

[4] User ID of owner

[5] Group ID of owner

[6] Device number

[7] Inode number

[8] Number of hard links

[9] Device type (if inode device) Major

[10] Device type (if inode device) Minor

[11] Time of last access

May be disabled in 'mount' with noatime

atime of files changed by exec, read, pipe, utime, mknod (mmap?)

atime of directories changed by addition/deletion of files

[12] Time of last modification

mtime of files changed by write, truncate, utime, mknod

mtime of directories changed by addtition/deletion of files

[13] Time of last change

ctime reflects time of changed inode information (owner, group

permissions, link count

-*-*- Per:

Return code: 0

Size of array: 14

Contents of array

Element 0: /home/mszick

Element 1: 4096

Element 2: 8

Element 3: 41e8

Element 4: 500

Element 5: 500

Element 6: 303

Element 7: 32385

Element 8: 22

Element 9: 0

Element 10: 0

Element 11: 1051221030

Element 12: 1051214068

Element 13: 1051214068

For a link in the form of linkname -> realname

stat -t linkname returns the linkname (link) information

stat -lt linkname returns the realname information

stat -tf and stat -ltf fields

[0] name

[1] ID-0? # Maybe someday, but Linux stat structure

[2] ID-0? # does not have either LABEL nor UUID

# fields, currently information must come

# from file-system specific utilities

These will be munged into:

[1] UUID if possible

[2] Volume Label if possible

Note: 'mount -l' does return the label and could return the UUID

[3] Maximum length of filenames

[4] Filesystem type

[5] Total blocks in the filesystem

[6] Free blocks

[7] Free blocks for non-root user(s)

[8] Block size of the filesystem

[9] Total inodes

[10] Free inodes

-*-*- Per:

Return code: 0

Size of array: 11

Contents of array

Element 0: /home/mszick

Element 1: 0

Element 2: 0

Element 3: 255

Element 4: ef53

Element 5: 2581445

Element 6: 2277180

Element 7: 2146050

Element 8: 4096

Element 9: 1311552

Element 10: 1276425

StatFieldsDoc

# LocateFile [-l] FileName Location-Array-Name

# LocateFile [-l] -of FileName Location-Array-FileName

LocateFile()

{

local -a LOC LOC1 LOC2

local lk="" of=0

case "$#" in

0) return 1 ;;

1) return 1 ;;

2) : ;;

*) while (( "$#" > 2 ))

do

case "$1" in

-l) lk=-1 ;;

-of) of=1 ;;

*) return 1 ;;

esac

shift

done ;;

esac

# More Sanscrit-2.0.5

# LOC1=( $(stat -t $lk $1) )

# LOC2=( $(stat -tf $lk $1) )

# Uncomment above two lines if system has "stat" command installed.

LOC=( ${LOC1[@]:0:1} ${LOC1[@]:3:11}

${LOC2[@]:1:2} ${LOC2[@]:4:1} )

case "$of" in

0) eval $2=\( \"\$\{LOC\[@\]\}\" \) ;;

1) echo "${LOC[@]}" > "$2" ;;

esac

return 0

# Which yields (if you are lucky, and have "stat" installed)

# -*-*- Location Discriptor -*-*-

# Return code: 0

# Size of array: 15

# Contents of array

# Element 0: /home/mszick 20th Century name

# Element 1: 41e8 Type and Permissions

# Element 2: 500 User

# Element 3: 500 Group

# Element 4: 303 Device

# Element 5: 32385 inode

# Element 6: 22 Link count

# Element 7: 0 Device Major

# Element 8: 0 Device Minor

# Element 9: 1051224608 Last Access

# Element 10: 1051214068 Last Modify

# Element 11: 1051214068 Last Status

# Element 12: 0 UUID (to be)

# Element 13: 0 Volume Label (to be)

# Element 14: ef53 Filesystem type

}

# And then there was some test code

ListArray() # ListArray Name

{

local -a Ta

eval Ta=\( \"\$\{$1\[@\]\}\" \)

echo

echo "-*-*- List of Array -*-*-"

echo "Size of array $1: ${#Ta[*]}"

echo "Contents of array $1:"

for (( i=0 ; i<${#Ta[*]} ; i++ ))

do

echo -e "\tElement $i: ${Ta[$i]}"

done

return 0

}

declare -a CUR_DIR

# For small arrays

ListDirectory "${PWD}" CUR_DIR

ListArray CUR_DIR

declare -a DIR_DIG

DigestFile CUR_DIR DIR_DIG

echo "The new \"name\" (checksum) for ${CUR_DIR[9]} is ${DIR_DIG[0]}"

declare -a DIR_ENT

# BIG_DIR # For really big arrays - use a temporary file in ramdisk

# BIG-DIR # ListDirectory -of "${CUR_DIR[11]}/*" "/tmpfs/junk2"

ListDirectory "${CUR_DIR[11]}/*" DIR_ENT

declare -a DIR_IDX

# BIG-DIR # IndexList -if "/tmpfs/junk2" DIR_IDX

IndexList DIR_ENT DIR_IDX

declare -a IDX_DIG

# BIG-DIR # DIR_ENT=( $(cat /tmpfs/junk2) )

# BIG-DIR # DigestFile -if /tmpfs/junk2 IDX_DIG

DigestFile DIR_ENT IDX_DIG

# Small (should) be able to parallize IndexList & DigestFile

# Large (should) be able to parallize IndexList & DigestFile & the assignment

echo "The \"name\" (checksum) for the contents of ${PWD} is ${IDX_DIG[0]}"

declare -a FILE_LOC

LocateFile ${PWD} FILE_LOC

ListArray FILE_LOC

exit 0

Stephane Chazelas демонстрирует возможность объектно ориентированного подхода к программированию в Bash-сценариях.