Change Pascal case in a string to snake case

Question

I have the following text:

SendNoticeMsg (api.post = "/test/SendNoticeMsg")
GenerateMsg (api.post = "/test/GenerateMsg")
GetUserLastAction (api.post = "/test/GetUserLastAction")

And I want to change the text to be:

SendNoticeMsg (api.post = "/test/send_notice_msg")
GenerateMsg (api.post = "/test/generate_msg")
GetUserLastAction (api.post = "/test/get_user_last_action")

Description: I just want to change the URL path to a valid underscore style, so the solution shouldn't change any other irrelevant characters.

I tried use the sed script:

sed -E 's/(\/test\/.*)([A-Z]).*\"/\1\2_\L/'

But it not works.

Answer 1

perl -wnE'
    @p = m{(.*/)(.*)"};                       # break up into parts
    @w = $p[-1] =~ /([A-Z][a-z0-9]*)/g;       # extract (PascalCase-ed) words
    $p[-1] = join("_", map { lc } @w).q{")};  # low-case them, join with _ 
    say @p
' input.txt

Or overwrite input "in-place" by changing switches to perl -i.bak -wnE'...'

This assumes that the words, after the initial, can only have [a-z0-9]; adjust if needed.

Answer 2

Another perl solution:

$ perl -pe '$i=0; s/(?=\/test\/)(\S+)/$s=$1;$s=~s!([A-Z])!$i++?"_".lc($1):lc($1)!ge;$s/ge ' underscore2.txt
SendNoticeMsg (api.post = "/test/send_notice_msg")
GenerateMsg (api.post = "/test/generate_msg")
GetUserLastAction (api.post = "/test/get_user_last_action")
$

Enhancing further.

$ perl -pe '$i=0; s/(?=\/[^\/]+\/)(\S+)/$s=$1;$s=~s!([A-Z])!$i++?"_".lc($1):lc($1)!ge;$s/ge ' underscore2.txt
SendNoticeMsg (api.post = "/test/send_notice_msg")
GenerateMsg (api.post = "/test/generate_msg")
GetUserLastAction (api.post = "/test/get_user_last_action")
$

Answer 3

perl -pe's{/test/\K[^"]*}{ lc( $& =~ s/\w\K(?=\p{Lu})/_/gr ) }e'

or even

perl -pe's{/test/\K[^"]*}{ lc( $& =~ s/\B(?=\p{Lu})/_/gr ) }e'

We extract the path, and replace it with the string constructed by the replacement expression. We construct that string by inserting underscores in the appropriate places (between word characters and uppercase letters), then lowercasing the result.

See Specifying file to process to Perl one-liner.

Answer 4

I'll add one more sed to the mix, a slight variation on the answer by @tripleee using # as an alternate substitution delimiter, e.g.

sed -E ':a;s#^(.*/)([^A-Z]*)([A-Z])#\1\L\2_\3#;ta;s#/_#/#'

The ^(.*/) allowing a greedy match of all characters from the beginning through the last '/' in each line.

Example Use/Output

$ sed -E ':a;s#^(.*/)([^A-Z]*)([A-Z])#\1\L\2_\3#;ta;s#/_#/#' file
SendNoticeMsg (api.post = "/test/send_notice_msg")
GenerateMsg (api.post = "/test/generate_msg")
GetUserLastAction (api.post = "/test/get_user_last_action")

(note: if not using GNU sed you will need to replace each ';' with -e to mark the separate expressions)

If for some reason your sed doesn't support ERE, then the "picket fence" like BRE would be:

sed ':a;s#^\(.*/\)\([^A-Z]*\)\([A-Z]\)#\1\L\2_\3#;ta;s#/_#/#'

Answer 5

Here's a Perl oneliner:

perl -pe 'sub to_snake { my $v = $_[0]; $v =~ s/([a-z])([A-Z])/$1_$2/g; return lc $v; } s@(/test/)(\w+)@$1.to_snake $2@e'

Note that you can change \w here to suit your needs, but should be sufficient for most purposes.

Breaking it down:

Perl's -p argument reads the input & writes to standard output; -e allows the program to be specified as a string on the command-line.

This part does the substitution on each line; it uses the e modifier to execute code:

s@(/test/)(\w+)@$1.to_snake $2@e'

This part defines the conversion function; it matches a lowercase letter preceding an uppercase letter, adds an underscore in between them and subsequently maps everything to lowercase using Perl's lc function:

sub to_snake { my $v = $_[0]; $v =~ s/([a-z])([A-Z])/$1_$2/g; return lc $v; }

Answer 6

Your simple sed script performs a single replacement. You want to add a loop to make it match and replace as many times as necessary.

sed -E -e ':a' -e 's/(\/test\/.*)([A-Z])(.*\")/\1_\L\2\3/' -e 'ta' -e 's%/_%/%'

The :a creates a label a and ta branches back to that label if the latest replacement was successful. (I'm blindly assuming the other parts of your script are correct for you; the availability of the -E option and whether it provides an escape \L for lowercasing is by no means universal or portable, though I had to refactor it a bit, and add a fix for /test/_whatever/.)

Answer 7

It seems all the answers use perl so here's one using awk

awk -F '"' '{
    gsub(/([^\/][A-Z])/," & ", $2);
    n=split($2,c," ");
    s=""
    for (i=1; i<=n; i++) {
        s = (i%2==1) ? (s c[i]) : (s substr(c[i],1,1) "_" substr(c[i], 2, 1));
    }
    $2=tolower(s);
    print $1 "\"" $2 "\"" $3;
}' input.txt

it's assuming no capturing groups or gensub available so it's separating the groups with spaces to later insert the _.

Answer 8

Assuming that you have 0 or 2 " per line and are interesting only in altering what is between them you might use AWK following way. Let file.txt content be

SendNoticeMsg (api.post = "/test/SendNoticeMsg")
GenerateMsg (api.post = "/test/GenerateMsg")
GetUserLastAction (api.post = "/test/GetUserLastAction")

then

BEGIN{FS=OFS="\""}{while(match($2,/[a-z][A-Z]/)){$2 = substr($2, 1, RSTART) "_" substr($2, RSTART+1)};$2=tolower($2);print}

output

SendNoticeMsg (api.post = "/test/send_notice_msg")
GenerateMsg (api.post = "/test/generate_msg")
GetUserLastAction (api.post = "/test/get_user_last_action")

Explanation I set both field seperator and output field seperator to ", then I alter $2 (what is inside ") as follows: as long as there is lowercase letter followed by upper case letter insert _ between them. Finally I change whole $2 to lowercase. (tested in gawk 4.2.1)

Answer 9

/in before lock

perl -pe '$_ = lc join "_", split /(?<=[^A-Z])(?=[A-Z])/;'

my $s = "Foo1BarXBazX";
print $s =~ s/^([A-Z])|(?<=[^A-Z])([A-Z]+)/defined $1 ? "\L$1" : "_\L$2"; /egr;
"foo1_bar_xbaz_x"