Improve include/exclude search in workspace #9307

alvsan09 · 2021-04-06T17:00:27Z

The following descriptions apply to the user input for the fields
'Files to include' and 'Files to exclude' for the 'Search in
workspace' feature.

Enhancements:

Handle provided relative paths e.g. ./README.md, README.md to only
apply the specified file.
Includes are limited to specific folder e.g. './test' should
not include './dir/path/test'.
Input strings (i.e. not resolving to relative paths) are converted
to two globs in order to include results of a resulting file and
include files under a resulting folder i.e. resolved globs are
**/string and **/string/*.
E.g. input file '.snaphot' shall include a file '.snapshot' or if
it's a folder it should apply to all the files underneath it.
'Files to include' can reference folders outside the workspace

Fixes: #8469

The search strategy offered by the search in workspace server
(RipgrepSearchInWorkspaceServer.search) is changed as follows:

By default, sets the search directories for the string WHAT to the provided ROOTURIS directories
and returns the assigned search id.

The include / exclude (options in SearchInWorkspaceOptions) are lists of patterns for files to
include / exclude during search (glob characters are allowed).

include patterns successfully recognized as absolute paths will override the default search and set
the search directories to the ones provided as includes.
Relative paths are allowed, the application will attempt to translate them to valid absolute paths
based on the applicable search directories.

Signed-off-by: Alvaro Sanchez-Leon alvaro.sanchez-leon@ericsson.com

What it does

For strings provided for files to include / exclude

Checks if the 'include' strings provided are valid paths to the available root folders and if so convert them to glob patterns
If the include strings are valid paths, replace the root paths with the resulting paths to be consumed by the 'ripgrep' executable.
Consider each string not resolving to sub-folders as a pattern to either a file or folder, therefore generating two glob entries for each.

How to test

4 new unit test cases provided

Manual testing can be performed as follows:

Using theia repo as an example directory structure:

Search to a relative file path of workspace root(s)
- Search pattern: cloud (produces a handful of matches)
- Enter 'Files to Include' text to: ./README.md (verify that only one file is considered)
'Files to include' can refer to files
- Search pattern: example
- Enter 'Files to include' to: .eslintcache (verify that multiple matches are reported)
'Files to include' can refer to folders
- Search pattern: example
- Enter 'Files to include' to: test (verify that multiple matches are reported)
Searching limited to a given subfolder
- Search pattern: example
- Enter 'Files to include' to: */doc/ (note that matches in two different folders are found i.e. 'doc' and 'packages/plugin-ext/doc'
- Replace 'Files to include' to doc (note that this now resolves matches for a single folder i.e. 'doc')
From the explorer view select a file or folder and from the context menu select 'Find in folder', press enter, and verify the results are as expected.
Using the 'Files to include' box, enter absolute paths to external folders, validate expected results.