I\'m developing a git post-receive hook in Python. Data is supplied on stdin
with lines similar to
ef4d4037f8568e386629457d4d960915a85da2ae 61a4
Let's dissect the various rules and build regex parts from them:
They can include slash /
for hierarchical (directory) grouping, but no slash-separated component can begin with a dot .
or end with the sequence .lock
.
# must not contain /.
(?!.*/\.)
# must not end with .lock
(?<!\.lock)$
They must contain at least one /
. This enforces the presence of a category like heads/, tags/ etc. but the actual names are not restricted. If the --allow-onelevel
option is used, this rule is waived.
.+/.+ # may get more precise later
They cannot have two consecutive dots ..
anywhere.
(?!.*\.\.)
They cannot have ASCII control characters (i.e. bytes whose values are lower than \040
, or \177 DEL
), space, tilde ~
, caret ^
, or colon :
anywhere.
[^\000-\037\177 ~^:]+ # pattern for allowed characters
They cannot have question-mark ?
, asterisk *
, or open bracket [
anywhere. See the --refspec-pattern
option below for an exception to this rule.
[^\000-\037\177 ~^:?*[]+ # new pattern for allowed characters
They cannot begin or end with a slash /
or contain multiple consecutive slashes (see the --normalize
option below for an exception to this rule)
^(?!/)
(?<!/)$
(?!.*//)
They cannot end with a dot .
.
(?<!\.)$
They cannot contain a sequence @{
.
(?!.*@\{)
They cannot contain a \
.
(?!.*\\)
Piecing it all together we arrive at the following monstrosity:
^(?!.*/\.)(?!.*\.\.)(?!/)(?!.*//)(?!.*@\{)(?!.*\\)[^\000-\037\177 ~^:?*[]+/[^\000-\037\177 ~^:?*[]+(?<!\.lock)(?<!/)(?<!\.)$
And if you want to exclude those that start with build-
then just add another lookahead:
^(?!build-)(?!.*/\.)(?!.*\.\.)(?!/)(?!.*//)(?!.*@\{)(?!.*\\)[^\000-\037\177 ~^:?*[]+/[^\000-\037\177 ~^:?*[]+(?<!\.lock)(?<!/)(?<!\.)$
This can be optimized a bit as well by conflating a few things that look for common patterns:
^(?!@$|build-|/|.*([/.]\.|//|@\{|\\))[^\000-\037\177 ~^:?*[]+/[^\000-\037\177 ~^:?*[]+(?<!\.lock|[/.])$
git check-ref-format <ref>
with subprocess.Popen
is a possibility:
import subprocess
process = subprocess.Popen(["git", "check-ref-format", ref])
exit_status = process.wait()
Advantages:
Disadvantages:
pygit2, which uses C bindings to libgit2, would be an even better possibility if check-ref-format
is exposed there, as it would be faster than Popen
, but I haven't found it.
For anyone coming to this question looking for the PCRE regular expression to match a valid Git branch name, it is the following:
^(?!/|.*([/.]\.|//|@\{|\\\\))[^\040\177 ~^:?*\[]+(?<!\.lock|[/.])$
This is an amended version of the regular expression written by Joey. In this version, however, an oblique is not required (it is for matching branchName
rather than refs/heads/branchName
).
Please refer to his correct answer to this question. He provides a full breakdown of each part of the regex, and how it relates to each requirement specified on the git-check-ref-format(1) manual pages.
If You want to check if reference is valid with pygit2 You can do like that function (code copied from documentation):
from pygit2 import reference_is_valid_name
reference_is_valid_name("refs/heads/master")
There's no need to write monstrosities in Perl. Just use /x:
# RegExp rules based on git-check-ref-format
my $valid_ref_name = qr%
^
(?!
# begins with
/| # (from #6) cannot begin with /
# contains
.*(?:
[/.]\.| # (from #1,3) cannot contain /. or ..
//| # (from #6) cannot contain multiple consecutive slashes
@\{| # (from #8) cannot contain a sequence @{
\\ # (from #9) cannot contain a \
)
)
# (from #2) (waiving this rule; too strict)
[^\040\177 ~^:?*[]+ # (from #4-5) valid character rules
# ends with
(?<!\.lock) # (from #1) cannot end with .lock
(?<![/.]) # (from #6-7) cannot end with / or .
$
%x;
foreach my $branch (qw(
master
.master
build/master
ref/HEAD/blah
/HEAD/blah
HEAD/blah/
master.lock
head/@{block}
master.
build//master
build\master
build\\master
),
'master blaster',
) {
print "$branch --> ".($branch =~ $valid_ref_name)."\n";
}
Joey++ for some of the code, though I made some corrections.
Taking the rules directly from the linked page, the following regular expression should match only valid branch names in refs/heads
not starting with "build-":
refs/heads/(?!.)(?!build-)((?!\.\.)(?!@{)[^\cA-\cZ ~^:?*[\\])+))(?<!\.)(?<!\.lock)
This starts with refs/heads
as yours does.
Then (?!build-)
checks that the next 6 characters are not build-
and (?!.)
checks that the branch does not start with a .
.
The entire group (((?!\.\.)(?!@{)[^\cA-\cZ ~^:?*[\\])+)
matches the branch name.
(?!\.\.)
checks that there are no instances of two periods in a row, and (?!@{)
checks that the branch does not contain @{
.
Then [^\cA-\cZ ~^:?*[\\]
matches any of the allowed characters by excluding control characters \cA-\cZ
and all of the rest of the characters that are specifically forbidden.
Finally, (?<!\.)
makes sure that the branch name did not end with a period and (?<!.lock)
checks that it did not end with .\lock
.
This can be extended to similarly match valid branch names in arbitrary folders, you can use
(?!.)((?!\.\.)(?!@{)[^\cA-\cZ ~^:?*[\\])+))(/(?!.)((?!\.\.)(?!@{)[^\cA-\cZ ~^:?*[\\])+)))*?/(?!.)(?!build-)((?!\.\.)(?!@{)[^\cA-\cZ ~^:?*[\\])+))(?<!\.)(?<!\.lock)
This applies basically the same rules to each piece of the branch name, but only checks that the last one does not start with build-