Age | Commit message (Collapse) | Author |
|
- Disabled gcc-6 since we develop with later tools anyway
- Turned the release test into a simple integration test
- Adding MacOSX on Travis-ci
Tests: Adding tests for Plink w. LOCO
Safety: Introduce strtok_safe to get rid of segfaults
|
|
Replaced eigenlib_dgemm with fast_dgemm - 10-30% speed gain for GEMMA
|
|
|
|
|
|
|
|
|
|
|
|
https://github.com/genetics-statistics/GEMMA/issues/81
|
|
|
|
|
|
https://github.com/genetics-statistics/GEMMA/issues/78
|
|
- Matrix checks as described in https://github.com/genetics-statistics/GEMMA/issues/72
- introduces -strict switch which will exit on certain conditions
- zero small eigenvalues in EigenDecomp_Zeroed which also checks for negative values
- commented out float versions of functions in lapack.cpp (pre-removal)
- reverted on disabled regression tests (GEMMA shows its previous behaviour now)
|
|
|
|
|
|
https://github.com/genetics-statistics/GEMMA/issues/58
|
|
|
|
debug.h: replaced assert_fail with enforce_fail
|
|
1 switch for LOCO of chromosome 1.
What are the use cases?
1. User runs vanilla GEMMA: all SNPs are considered input for GWA and K
2. User passes in -snps: all these SNPs are considered for GWA and K
3. User passes in -snps and -ksnps: All these SNPs are used for GWA,
Ksnps are used for K
4. User passes in -loco: SNPs are split by chromosome (GWA incl., K
excl.)
5. User passes in -snps, -gwasnps and -ksnps could mean that also GWA
is subset explicitely (nyi)
In all cases indicator_snp is honored and we get the most flexible way for
studying SNP combinations that can be passed in in different ways.
Overall added:
- various comments in source code
- tests in test framework inlc. fast-check
- NDEBUG compilation support in the Makefile
- -debug switch for GEMMA debug output
- debug.h which includes enforce functions which work like
assert. Unlike assert, enforce also works in release compilation
- -nind switch limit the number of individuals used
(trim_individuals for testing)
- enforcing tests of input files - e.g. are number of individuals correct
- checks for memory allocation - we should add more of those
- more checks for gsl results - we should add more of those
- replaced strtoken with regex as a first case. They should all be
replaced. strtoken is not thread safe, for one.
- introduced C++ iterators
- introduced C++ closure in BimBam LMM for cached processing
- more localized initialization of variables - makes for demonstratably
more correct code
- -ksnps adds snps into setKSnps
- -gwasnps adds snps into setGWASnps
- both sets are computed by -loco
- attempted to make the code easier to read
|
|
|
|
|
|
shunit2 in repo
fixes https://github.com/genetics-statistics/GEMMA/issues/55
|
|
|
|
|
|
|
|
tests passed: 13
tests failed: 0
tests total: 13
success rate: 100%
|
|
|
|
|
|
|
|
Also added 'make check' command in Makefile and documented in INSTALL.md
|
|
|
|
|