I'm trying to compute the source code size of gcc by considering cpp files first:

# NOTE: the cpp loop finishes immediately
LOC=0
BYTES=0
FILES=$(find . -name "*.cpp")
for f in ${FILES};
do
    BYTES_TAG=$(stat --printf="%d" $f)
    LOC_TAG=$(cat $f | wc -l)
    BYTES=$((BYTES BYTES_TAG))
    LOC=$((LOC LOC_TAG))
done
echo "LOC = $LOC, SIZE = $BYTES"

Then I try to sum the *.c files, but the bash loop doesn't stop. here is my gcc version:

$ wget https://ftp.gnu.org/gnu/gcc/gcc-11.2.0/gcc-11.2.0.tar.gz
$ tar -xf gcc-11.2.0.tar.gz

This is weird because counting all the files is immediate with:

$ find . -type f | wc -l

CodePudding user response：

Size of all *.c and *.cpp files in bytes:

find . -name *.cpp -o -name *.c -exec wc -c {} \; | sed "s/ .*//" | paste -sd  | bc

Number of lines in all *.c and *.cpp files:

find . -name *.cpp -o -name *.c -exec wc -l {} \; | sed "s/ .*//" | paste -sd  | bc

Explanation: find ... -exec executes a command on all files it finds, replacing the {} in the -exec part with the file name(s). If you end the -exec part with \;, it will be executed once for each file. In some cases, ending the -exec part with is more efficient -- this will execute for as many files as will fit in one command line.

wc -l and wc -c will print one line per file, with the number of lines / characters (bytes) followed by the file name. The `sed "s/ .*//"' will cut away everything after the number.

paste -sd will then concatenate all the lines (-s) and separate them with plus signs (-d ). Piping this to bc makes bc execute the addition and give you the total you are looking for.

Meta answer: Learn about find ... -exec. Don't loop over find output, because you will get into trouble e.g. when file names contain spaces -- and xargs is in most cases unnecessary.

CodePudding user response：

Then I try to sum the *.c files, but the bash loop doesn't stop

You just waited not enough time. Bash is a very slow programming language. For every single loop, your program forks a subshell and does fork exec of stat, cat and wc. And you are reading each file twice - once by stat, than by cat. That's a lot of processes and a lot of work and double the needed I/O.

Write the same code in AWK, Python or Perl or in C or in C , it will be much faster.

This is weird because counting all the files is immediate with:

The code you posted is counting the bytes of files names, not files content. To count bytes of files content, you would do find ... | xargs wc ... - pass filenames as arguments to wc.

Yes, a single pipeline of constant count of two processes will be a lot faster. Opening of files, counting bytes and lines, and the sum is calculated inside a C program wc, not by Bash.

CodePudding user response：

Here is the TLDR answer based on @HolyBlackCat (see also here):

• install cloc with sudo apt install cloc
• from the top level gcc directory simply run: $ cloc .
• Then just look at the relevant info (C/CPP for me):

------------------------------------------------------------
Language  files          blank        comment         code
------------------------------------------------------------
C         45147         680546         889750        3465801
C         27752         230223         289596        1097375