complex search/delete/move/replace operation using sed? - regex

Question

after several hours of searching and experimenting, I'm hoping someone can either help me or rub my nose in a post I've missed which acctually would be helpful as well come to think of it...
Problem:
I've made a quick&dirty fix in several dozens of php scripts (we use to enhance smarty capabilities) with security checks.
Example of input(part1):
///// SMARTY AUTH /////
$auth['model'] = isset($params['model']) ? $params['model'] : null;
$auth['requiredLevel'] = isset($params['requiredlevel']) ? $params['requiredlevel'] : null;
$auth['baseAuthorizationLevel'] = isset($params['_authorizationlevel']) ? $params['_authorizationlevel'] : null;
$auth['defaultRequiredLevel'] = AuthorizationLevel::AULE_WRITE;
$auth['baseModel'] = $smarty->getTemplateVars('model');
///// SMARTY AUTH /////
...which i'd like to replace with a much cleaner solution we've come up with. Now here's the rub; in one section of the file there's a block of lines, luckily with very distinct delimiter lines, but in one of those lines is a piece of code that needs to be merged with a replacement string which replaces a second pattern in a line which follows the before-said block, with optionally a variable number of lines in between.
I'm having trouble figuring out how to piece this nested code together as the shorthand code of sed is quite confusing to me.
So far I've tried to assemble the code needed to capture the first block, but sed keeps giving me the same error each time; extra characters after command
here are some of the attempts I've made:
sed -n 'p/^\/\/\/\/\/ SMARTY AUTH \/\/\/\/\/\\n.*\\n.*\\n.*\\n.*AULE_\([A-Z_]*\);$^.*$^^\/\/\/\/\/ SMARTY AUTH \/\/\/\/\/$/' function.xls_form.php
sed -n 'p/\(^.*SMARTY AUTH.*$^.*$^.*$^.*$^.*AULE_\([A-Z_]*\);$^.*$^.*SMARTY AUTH.*$/' function.xls_form.php
the second part is relatively easy compared to the first;
sed -ei'.orig' 's/RoleContextAuthorizations::smartyAuth(\$auth)/$smarty->hasAccess(\$params,AuthorizationLevel::AULE_\1)/' *.php
where \1 would be the matched snippet from the first part...
Edit:
The first codeblock is an example of input part 1 which needs to be removed; part 2 is RoleContextAuthorizations::smartyAuth($auth) which needs to be replaced with $smarty->hasAccess($params, AuthorizationLevel::AULE_<snippet from part1>)
/edit
Hoping somebody can point me in the right direction, Many thanks in advance!!!

Answer 1

The hold space is going to be key to solving this. You can copy material from the pattern space (where sed normally works) into the hold space, and do various operations with the hold space, etc.
You need to find the AuthorizationLevel::AULE_WRITE type text within the block markers, and copy that to the hold space, and then delete the text within the block markers. And then separately find the other pattern and replace it with information from the hold space.
Given that the markers use slashes, it is also time to use a custom search marker which is introduced by a backslash. The following could be in a file script.sed, to be used as:
sed -f script.sed function.xls_form.php
When you're sure it's working, you can play with -i options to overwrite the original.
\%///// SMARTY AUTH /////%,\%///// SMARTY AUTH /////% {
/.*\(AuthorizationLevel::AULE_[A-Z]\{1,\}\).*/{
s//$smarty->hasAccess($params,\1);/
x
}
d
}
/RoleContextAuthorizations::smartyAuth($auth)/x
The first line searches for the start and end marker, using \% to change the delimiter to %. There's then a group of actions in braces. The second line searches for the authorization level and starts a second group of actions. The substitute command replaces the line with the desired output line. The x swaps the pattern space and the hold space, copying the desired output line to the hold space (and copying the empty hold space to the pattern space — it's x for eXchange pattern and hold spaces). This has saved the AuthorizationLevel information. The inner block ends; the outer block deletes the line and continues the execution. Note that there's no need to escape the $ symbol most of the time — it would matter if it was at the end of a pattern (there's a difference between /a\$/ and /a$/, but no difference between /b$c/ and /b\$c/).
The last line then looks for the RoleContextAuthorizations line and swaps it with the hold space. Everything else is just let through.
Given a data file containing:
Gibberish
Rhubarb
///// SMARTY AUTH /////
$auth['model'] = isset($params['model']) ? $params['model'] : null;
$auth['requiredLevel'] = isset($params['requiredlevel']) ? $params['requiredlevel'] : null;
$auth['baseAuthorizationLevel'] = isset($params['_authorizationlevel']) ? $params['_authorizationlevel'] : null;
$auth['defaultRequiredLevel'] = AuthorizationLevel::AULE_WRITE;
$auth['baseModel'] = $smarty->getTemplateVars('model');
///// SMARTY AUTH /////
More gibberish
More rhubarb - it is good with strawberries, especially in yoghurt
RoleContextAuthorizations::smartyAuth($auth);
Trailing gibbets — ugh; worse are trailing giblets
Finish - EOF
The output from sed -f script.sed data is:
$ sed -f script.sed data
Gibberish
Rhubarb
More gibberish
More rhubarb - it is good with strawberries, especially in yoghurt
$smarty->hasAccess($params,AuthorizationLevel::AULE_WRITE);
Trailing gibbets — ugh; worse are trailing giblets
Finish - EOF
$
I think that's what was wanted.
You can convert the file of sed script into a single line of gibberish, but that's left as an exercise for the reader — it isn't very hard, but GNU sed and BSD (macOS) sed have different rules for when you need semicolons as part of a single line command; you were warned. There are also differences in the rules for the -i option between the GNU and BSD variants of sed.
If you have to preserve some portions of the RoleContextAuthorizations::smartyAuth line, you have to work harder, but it can probably be done. For example, you can add the hold space to the current pattern space with the G command, and then edit the information into the right places. It is simplest if every place the line occurs needs to look the same apart from the AULE_XYZ string — that's what I've assumed here.
Also, note that using x rather than h or g is lazy — but doesn't matter if there's only one RoleContextAuthorizations::smartyAuth line. Using the alternatives would mean that if a file has multiple RoleContextAuthorizations::smartyAuth lines, then you'd be able to make the same substitution in each, unless there's another ///// SMARTY AUTH ///// in the file.