When using expand_vars=True, and the content is like "${VAR}ANOTHER_CONTENT", the ANOTHER_CONTENT will be ignored.

[flymanzhao]

code

import os
os.environ["B"] = "${A};Val_B;"
os.environ["A"] = "Val_A"
from environs import Env
env = Env(expand_vars=True)
print(env("B"))

expect:

Val_A;Val_B;

what I got:

Val_A

is this a bug? Thanks very much!

[sloria]

Hm, I'd also expect expanded variables to work with other characters included. Would def appreciate a PR if any kind soul looks into this

[basedBaba]

Hello! I am a first-time contributor to environs and new to open-source in general, so I apologize in advance if something I say does not make sense.

This bug seems to be caused by expand_match, which only matches and expands when the variable is present at the beginning and ignores the rest.

environs/src/environs/__init__.py

Lines 542 to 556 in ddd66d3

	expand_match = self.expand_vars and _EXPANDED_VAR_PATTERN.match(value)
	if expand_match: # Full match expand_vars - special case keep default
	proxied_key: _StrType = expand_match.groups()[0]
	subs_default: typing.Optional[_StrType] = expand_match.groups()[1]
	if subs_default is not None:
	default = subs_default[2:]
	elif (
	value == default
	): # if we have used default, don't use it recursively
	default = ma.missing
	return (
	key,
	self._get_from_environ(proxied_key, default, proxied=True)[1],
	proxied_key,
	)

Meanwhile, expand_search along with _expand_vars seems to handle all cases, whether the variable is present at the beginning or in the middle, and also seems to handle default values as well.

environs/src/environs/__init__.py

Lines 557 to 561 in ddd66d3

	expand_search = self.expand_vars and _EXPANDED_VAR_PATTERN.search(value)
	if (
	expand_search
	): # Multiple or in text match expand_vars - General case - default lost
	return self._expand_vars(env_key, value)

Using expand_search for all cases instead of expand_match seems to fix this particular bug, not sure if it breaks some other functionality.

Please let me know if this approach makes sense, so I can open a pull request for the same. Thank you!

[flymanzhao]

Thanks @basedBaba , I also noticed this. I think you are correct. But I also want to point out that the r"$" is not handled correctly either. We'd better handle the var expansion and the r"$" replacement in one time.