Version
codebase-memory-mcp 0.10.8
Platform
Linux (WSL2, x86_64)
Install channel
GitHub release archive / install.sh
Binary variant
portable
What happened, and what did you expect?
The README documents name_pattern as a regex (--name-pattern '.*Handler.*'), so I assumed file_pattern was one too. It isn't. It's a glob, and a regex pattern returns 0 results with no error.
--label Function -> total: 13006
--label Function --file-pattern '.*src.*' -> total: 0
--label Function --file-pattern 'src/*' -> total: 2162
Poking at it, the glob looks restricted. * and ? both work and cross /, [a-z] matches nothing, and matching is case insensitive: SRC/* returns the same 2162 as src/*.
tools/list returns an empty description for file_pattern, name_pattern and qn_pattern, so there's nowhere to look this up short of guessing.
Documenting the syntax would be enough. Rejecting patterns that contain regex-only characters would be better, since a zero here looks exactly like a real "nothing matched".
Version
codebase-memory-mcp 0.10.8
Platform
Linux (WSL2, x86_64)
Install channel
GitHub release archive / install.sh
Binary variant
portable
What happened, and what did you expect?
The README documents
name_patternas a regex (--name-pattern '.*Handler.*'), so I assumedfile_patternwas one too. It isn't. It's a glob, and a regex pattern returns 0 results with no error.Poking at it, the glob looks restricted.
*and?both work and cross/,[a-z]matches nothing, and matching is case insensitive:SRC/*returns the same 2162 assrc/*.tools/listreturns an emptydescriptionforfile_pattern,name_patternandqn_pattern, so there's nowhere to look this up short of guessing.Documenting the syntax would be enough. Rejecting patterns that contain regex-only characters would be better, since a zero here looks exactly like a real "nothing matched".