biology/vcflib: C++ library and CLI tools for parsing and manipulating VCF files

Approved by:    jrm (mentor)
Differential Revision:  https://reviews.freebsd.org/D15109

7 files changed, 510 insertions, 0 deletions

diff --git a/biology/vcflib/Makefile b/biology/vcflib/Makefile
new file mode 100644
index 000000000000..ffec9fe35272
--- /dev/null
+++ b/biology/vcflib/Makefile
@@ -0,0 +1,47 @@
+# $FreeBSD$
+
+PORTNAME=		vcflib
+DISTVERSIONPREFIX=	v
+DISTVERSION=		1.0.0-rc1-130
+DISTVERSIONSUFFIX=	-g7e3d806
+CATEGORIES=		biology
+
+MAINTAINER=	jwb@FreeBSD.org
+COMMENT=	C++ library and CLI tools for parsing and manipulating VCF files
+
+LICENSE=	MIT
+
+LIB_DEPENDS=	libhts.so:biology/htslib \
+		libtabix.so:biology/tabixpp \
+		libsw.so:biology/smithwaterman
+COMMON_DEPENDS=	fastahack:biology/fastahack \
+		filevercmp:sysutils/filevercmp \
+		fsom:science/fsom \
+		multichoose>=1.0.3:math/multichoose \
+		interval_tree_test:math/intervaltree
+BUILD_DEPENDS=	${COMMON_DEPENDS}
+RUN_DEPENDS=	${COMMON_DEPENDS}
+
+USES=		gmake shebangfix
+USE_LDCONFIG=	yes
+USE_GITHUB=	yes
+GH_ACCOUNT=	ekg
+
+CXXFLAGS+=	-I${LOCALBASE}/include/smithwaterman \
+		-I${LOCALBASE}/include/multichoose \
+		-I${LOCALBASE}/include/filevercmp \
+		-I${LOCALBASE}/include/fastahack \
+		-I${LOCALBASE}/include/intervaltree \
+		-DVERSION='\"${PORTVERSION}\"'
+
+# Clang and GCC disable sse2 by default on i386, but it's required for vcflib
+CFLAGS_i386=	-msse2
+
+MAKEFILE=	${FILESDIR}/Makefile.external-libs
+INSTALL_TARGET=	install-strip
+
+post-install:
+	${RLN} ${STAGEDIR}${PREFIX}/lib/libvcflib.so.1 \
+		${STAGEDIR}${PREFIX}/lib/libvcflib.so
+
+.include <bsd.port.mk>
diff --git a/biology/vcflib/distinfo b/biology/vcflib/distinfo
new file mode 100644
index 000000000000..985916a8d85b
--- /dev/null
+++ b/biology/vcflib/distinfo
@@ -0,0 +1,3 @@
+TIMESTAMP = 1523243322
+SHA256 (ekg-vcflib-v1.0.0-rc1-130-g7e3d806_GH0.tar.gz) = 8ec24354dfc7a87c011c865ab75f3eb72646768ab9cb6eff4263e69763338478
+SIZE (ekg-vcflib-v1.0.0-rc1-130-g7e3d806_GH0.tar.gz) = 20138644
diff --git a/biology/vcflib/files/Makefile.external-libs b/biology/vcflib/files/Makefile.external-libs
new file mode 100644
index 000000000000..ae030a517b44
--- /dev/null
+++ b/biology/vcflib/files/Makefile.external-libs
@@ -0,0 +1,207 @@
+#OBJ_DIR = ./
+HEADERS = src/Variant.h \
+		  src/split.h \
+		  src/pdflib.hpp \
+		  src/var.hpp \
+                  src/cdflib.hpp \
+		  src/rnglib.hpp \
+		  src/join.h
+SOURCES = src/Variant.cpp \
+		  src/rnglib.cpp \
+		  src/var.cpp \
+		  src/pdflib.cpp \
+		  src/cdflib.cpp \
+		  src/split.cpp
+OBJECTS= $(SOURCES:.cpp=.o)
+
+VCF_LIB_LOCAL:=$(shell pwd)
+BIN_DIR:=bin
+LIB_DIR:=lib
+SRC_DIR=src
+INC_DIR:=include
+OBJ_DIR:=obj
+
+LIB =		libvcflib.a
+SOVERSION =	1
+SLIB =		libvcflib.so.$(SOVERSION)
+
+# TODO
+#vcfstats.cpp
+
+BIN_SOURCES = src/vcfecho.cpp \
+			  src/vcfnormalizesvs.cpp \
+			  src/dumpContigsFromHeader.cpp \
+			  src/bFst.cpp \
+			  src/pVst.cpp \
+			  src/hapLrt.cpp \
+			  src/popStats.cpp \
+			  src/wcFst.cpp \
+			  src/iHS.cpp \
+			  src/segmentFst.cpp \
+			  src/segmentIhs.cpp \
+			  src/genotypeSummary.cpp \
+			  src/sequenceDiversity.cpp \
+			  src/pFst.cpp \
+			  src/smoother.cpp \
+			  src/LD.cpp \
+			  src/plotHaps.cpp \
+			  src/abba-baba.cpp \
+			  src/permuteGPAT++.cpp \
+			  src/permuteSmooth.cpp \
+			  src/normalize-iHS.cpp \
+			  src/meltEHH.cpp \
+			  src/vcfaltcount.cpp \
+			  src/vcfhetcount.cpp \
+			  src/vcfhethomratio.cpp \
+			  src/vcffilter.cpp \
+			  src/vcf2tsv.cpp \
+			  src/vcfgenotypes.cpp \
+			  src/vcfannotategenotypes.cpp \
+			  src/vcfcommonsamples.cpp \
+			  src/vcfremovesamples.cpp \
+			  src/vcfkeepsamples.cpp \
+			  src/vcfsamplenames.cpp \
+			  src/vcfgenotypecompare.cpp \
+			  src/vcffixup.cpp \
+			  src/vcfclassify.cpp \
+			  src/vcfsamplediff.cpp \
+			  src/vcfremoveaberrantgenotypes.cpp \
+			  src/vcfrandom.cpp \
+			  src/vcfparsealts.cpp \
+			  src/vcfstats.cpp \
+			  src/vcfflatten.cpp \
+			  src/vcfprimers.cpp \
+			  src/vcfnumalt.cpp \
+			  src/vcfcleancomplex.cpp \
+			  src/vcfintersect.cpp \
+			  src/vcfannotate.cpp \
+			  src/vcfallelicprimitives.cpp \
+			  src/vcfoverlay.cpp \
+			  src/vcfaddinfo.cpp \
+			  src/vcfkeepinfo.cpp \
+			  src/vcfkeepgeno.cpp \
+			  src/vcfafpath.cpp \
+			  src/vcfcountalleles.cpp \
+			  src/vcflength.cpp \
+			  src/vcfdistance.cpp \
+			  src/vcfrandomsample.cpp \
+			  src/vcfentropy.cpp \
+			  src/vcfglxgt.cpp \
+			  src/vcfroc.cpp \
+			  src/vcfcheck.cpp \
+			  src/vcfstreamsort.cpp \
+			  src/vcfuniq.cpp \
+			  src/vcfuniqalleles.cpp \
+			  src/vcfremap.cpp \
+			  src/vcf2fasta.cpp \
+			  src/vcfsitesummarize.cpp \
+			  src/vcfbreakmulti.cpp \
+			  src/vcfcreatemulti.cpp \
+			  src/vcfevenregions.cpp \
+			  src/vcfcat.cpp \
+			  src/vcfgenosummarize.cpp \
+			  src/vcfgenosamplenames.cpp \
+			  src/vcfgeno2haplo.cpp \
+			  src/vcfleftalign.cpp \
+			  src/vcfcombine.cpp \
+			  src/vcfgeno2alleles.cpp \
+			  src/vcfindex.cpp \
+			  src/vcf2dag.cpp \
+			  src/vcfsample2info.cpp \
+			  src/vcfqual2info.cpp \
+			  src/vcfinfo2qual.cpp \
+			  src/vcfglbound.cpp \
+			  src/vcfunphase.cpp \
+			  src/vcfnull2ref.cpp \
+			  src/vcfinfosummarize.cpp
+
+# when we can figure out how to build on mac
+# src/vcfsom.cpp
+
+#BINS = $(BIN_SOURCES:.cpp=)
+BINS = $(addprefix bin/,$(notdir $(BIN_SOURCES:.cpp=)))
+SHORTBINS = $(notdir $(BIN_SOURCES:.cpp=))
+# Use ?= to allow overriding from the env or command-line.
+
+MAKE ?=		make
+LOCALBASE ?=	/usr/local
+LIB_PATH ?=	${LOCALBASE}/lib
+
+CC ?=		cc
+CXX ?=		c++
+CXXFLAGS ?=	-O3
+CFLAGS +=	-D_FILE_OFFSET_BITS=64 -fPIC
+CXXFLAGS +=	$(CFLAGS) --std=c++11
+#CXXFLAGS +=	-pedantic -Wall -Wshadow -Wpointer-arith -Wcast-qual
+
+DESTDIR ?=	stage
+PREFIX ?=	/usr/local
+STRIP ?=	strip
+INSTALL ?=	install -c
+MKDIR ?=	mkdir -p
+AR ?=		ar
+
+SSW =		src/ssw.o src/ssw_cpp.o
+
+INCLUDES =	-I${LOCALBASE}/include \
+		-I${LOCALBASE}/include/smithwaterman \
+		-I${LOCALBASE}/include/multichoose \
+		-I${LOCALBASE}/include/fastahack \
+		-I${LOCALBASE}/include/intervaltree
+LDFLAGS +=	-L. -lvcflib \
+		-L$(LIB_PATH) -lsw -ltabix -lhts -lfastahack -lfilevercmp \
+		-lpthread -lz -lm
+
+all: $(OBJECTS) $(BINS) $(LIB) $(SLIB)
+
+ssw.o:		src/ssw.h
+ssw_cpp.o:	src/ssw_cpp.h
+
+openmp:
+	$(MAKE) CXXFLAGS="$(CXXFLAGS) -fopenmp -D HAS_OPENMP"
+
+profiling:
+	$(MAKE) CXXFLAGS="$(CXXFLAGS) -g" all
+
+gprof:
+	$(MAKE) CXXFLAGS="$(CXXFLAGS) -pg" all
+
+$(OBJECTS): $(SOURCES) $(HEADERS)
+	$(CXX) -c -o $@ src/$(*F).cpp $(INCLUDES) $(CXXFLAGS)
+
+$(SHORTBINS):
+	$(MAKE) $(BIN_DIR)/$@
+
+$(BINS): $(BIN_SOURCES) $(LIB) $(OBJECTS) $(SSW) pre
+	$(CXX) src/$(notdir $@).cpp -o $@ $(INCLUDES) $(CXXFLAGS) $(LDFLAGS)
+
+$(LIB): $(OBJECTS) $(SSW)
+	ar rs $(LIB) $(OBJECTS) $(SSW)
+
+$(SLIB): $(OBJECTS) $(SSW)
+	$(CXX) -shared -Wl,-soname,$(SLIB) -o $(SLIB) $(OBJECTS) $(SSW)
+
+install: all
+	$(MKDIR) $(DESTDIR)$(PREFIX)/bin
+	$(MKDIR) $(DESTDIR)$(PREFIX)/include/vcflib
+	$(MKDIR) $(DESTDIR)$(PREFIX)/lib
+	$(INSTALL) bin/* $(DESTDIR)$(PREFIX)/bin
+	$(INSTALL) src/*.h src/*.hpp $(DESTDIR)$(PREFIX)/include/vcflib
+	$(INSTALL) $(LIB) $(SLIB) $(DESTDIR)$(PREFIX)/lib
+
+install-strip: install
+	$(STRIP) $(DESTDIR)$(PREFIX)/bin/* $(DESTDIR)$(PREFIX)/lib/$(SLIB)
+
+test: $(BINS)
+	@prove -Itests/lib -w tests/*.t
+
+clean:
+	rm -f $(BINS) $(OBJECTS)
+	rm -f ssw_cpp.o ssw.o
+	rm -f $(LIB)
+	rm -rf $(BIN_DIR)
+
+pre:
+	mkdir -p $(BIN_DIR)
+
+.PHONY: clean all test pre
diff --git a/biology/vcflib/files/Makefile.submod b/biology/vcflib/files/Makefile.submod
new file mode 100644
index 000000000000..aafad2809a9e
--- /dev/null
+++ b/biology/vcflib/files/Makefile.submod
@@ -0,0 +1,117 @@
+#OBJ_DIR = ./
+HEADERS = src/Variant.h \
+		  src/split.h \
+		  src/join.h
+SOURCES = src/Variant.cpp \
+		  src/split.cpp
+OBJECTS= $(SOURCES:.cpp=.o)
+
+VCF_LIB_LOCAL:=$(shell pwd)
+BIN_DIR:=bin
+LIB_DIR:=lib
+SRC_DIR=src
+INC_DIR:=include
+OBJ_DIR:=obj
+
+include Makefile.common
+
+TABIX = tabixpp/tabix.o
+FASTAHACK = fastahack/Fasta.o
+SMITHWATERMAN = smithwaterman/SmithWatermanGotoh.o
+REPEATS = smithwaterman/Repeats.o
+INDELALLELE = smithwaterman/IndelAllele.o
+DISORDER = smithwaterman/disorder.o
+LEFTALIGN = smithwaterman/LeftAlign.o
+FSOM = fsom/fsom.o
+FILEVERCMP = filevercmp/filevercmp.o
+
+INCLUDES = -Itabixpp/htslib -I$(INC_DIR) -L. -Ltabixpp/htslib
+LDFLAGS = -L$(LIB_DIR) -lvcflib -lhts -lpthread -lz -lm
+
+
+all: $(OBJECTS) $(BINS)
+
+CXX ?= c++
+CXXFLAGS ?= -O3 -D_FILE_OFFSET_BITS=64
+#CXXFLAGS = -O2
+#CXXFLAGS = -pedantic -Wall -Wshadow -Wpointer-arith -Wcast-qual
+
+SSW = src/ssw.o src/ssw_cpp.o
+
+ssw.o: src/ssw.h
+ssw_cpp.o:src/ssw_cpp.h
+
+openmp:
+	$(MAKE) CXXFLAGS="$(CXXFLAGS) -fopenmp -D HAS_OPENMP"
+
+profiling:
+	$(MAKE) CXXFLAGS="$(CXXFLAGS) -g" all
+
+gprof:
+	$(MAKE) CXXFLAGS="$(CXXFLAGS) -pg" all
+
+$(OBJECTS): $(SOURCES) $(HEADERS) $(TABIX) multichoose pre $(SMITHWATERMAN) $(FILEVERCMP)
+	$(CXX) -c -o $@ src/$(*F).cpp $(INCLUDES) $(LDFLAGS) $(CXXFLAGS) && cp src/*.h* $(VCF_LIB_LOCAL)/$(INC_DIR)/
+
+multichoose: pre
+	cd multichoose && $(MAKE) && cp *.h* $(VCF_LIB_LOCAL)/$(INC_DIR)/
+
+intervaltree: pre
+	cd intervaltree && $(MAKE) && cp *.h* $(VCF_LIB_LOCAL)/$(INC_DIR)/
+
+$(TABIX): pre
+	cd tabixpp && $(MAKE) && cp *.h* $(VCF_LIB_LOCAL)/$(INC_DIR)/
+
+$(SMITHWATERMAN): pre
+	cd smithwaterman && $(MAKE) && cp *.h* $(VCF_LIB_LOCAL)/$(INC_DIR)/ && cp *.o $(VCF_LIB_LOCAL)/$(OBJ_DIR)/
+
+$(DISORDER): $(SMITHWATERMAN)
+
+$(REPEATS): $(SMITHWATERMAN)
+
+$(LEFTALIGN): $(SMITHWATERMAN)
+
+$(INDELALLELE): $(SMITHWATERMAN)
+
+$(FASTAHACK): pre
+	cd fastahack && $(MAKE) && cp *.h* $(VCF_LIB_LOCAL)/$(INC_DIR)/ && cp Fasta.o $(VCF_LIB_LOCAL)/$(OBJ_DIR)/
+
+#$(FSOM):
+#	cd fsom && $(CXX) $(CXXFLAGS) -c fsom.c -lm
+
+$(FILEVERCMP): pre
+	cd filevercmp && make && cp *.h* $(VCF_LIB_LOCAL)/$(INC_DIR)/ && cp *.o $(VCF_LIB_LOCAL)/$(INC_DIR)/
+
+$(SHORTBINS): pre
+	$(MAKE) bin/$@
+
+$(BINS): $(BIN_SOURCES) libvcflib.a $(OBJECTS) $(SMITHWATERMAN) $(FASTAHACK) $(DISORDER) $(LEFTALIGN) $(INDELALLELE) $(SSW) $(FILEVERCMP) pre intervaltree
+	$(CXX) src/$(notdir $@).cpp -o $@ $(INCLUDES) $(LDFLAGS) $(CXXFLAGS)
+
+libvcflib.a: $(OBJECTS) $(SMITHWATERMAN) $(REPEATS) $(FASTAHACK) $(DISORDER) $(LEFTALIGN) $(INDELALLELE) $(SSW) $(FILEVERCMP) $(TABIX) pre
+	ar rs libvcflib.a $(OBJECTS) smithwaterman/sw.o $(FASTAHACK) $(SSW) $(FILEVERCMP) $(TABIX)
+	cp libvcflib.a $(LIB_DIR)
+
+
+test: $(BINS)
+	@prove -Itests/lib -w tests/*.t
+
+pre:
+	if [ ! -d $(BIN_DIR) ]; then mkdir -p $(BIN_DIR); fi
+	if [ ! -d $(LIB_DIR) ]; then mkdir -p $(LIB_DIR); fi
+	if [ ! -d $(INC_DIR) ]; then mkdir -p $(INC_DIR); fi
+	if [ ! -d $(OBJ_DIR) ]; then mkdir -p $(OBJ_DIR); fi
+
+clean:
+	rm -f $(BINS) $(OBJECTS)
+	rm -f ssw_cpp.o ssw.o
+	rm -f libvcflib.a
+	rm -rf $(BIN_DIR)
+	rm -rf $(LIB_DIR)
+	rm -rf $(INC_DIR)
+	rm -rf $(OBJ_DIR)
+	cd tabixpp && make clean
+	cd smithwaterman && make clean
+	cd fastahack && make clean
+
+.PHONY: clean all test pre
diff --git a/biology/vcflib/files/patch-src_cdflib.cpp b/biology/vcflib/files/patch-src_cdflib.cpp
new file mode 100644
index 000000000000..84bff432c1ed
--- /dev/null
+++ b/biology/vcflib/files/patch-src_cdflib.cpp
@@ -0,0 +1,11 @@
+--- src/cdflib.cpp.orig	2018-03-09 20:31:19 UTC
++++ src/cdflib.cpp
+@@ -10040,7 +10040,7 @@ void negative_binomial_cdf_values ( int 
+     1, 2, 3,
+     0, 1, 2 };
+ 
+-  if ( n_data < 0 )
++  if ( *n_data < 0 )
+   {
+     *n_data = 0;
+   }
diff --git a/biology/vcflib/pkg-descr b/biology/vcflib/pkg-descr
new file mode 100644
index 000000000000..977e86a03dae
--- /dev/null
+++ b/biology/vcflib/pkg-descr
@@ -0,0 +1,21 @@
+The Variant Call Format (VCF) is a flat-file, tab-delimited textual format
+intended to concisely describe reference-indexed variations between
+individuals. VCF provides a common interchange format for the description of
+variation in individuals and populations of samples, and has become the defacto
+standard reporting format for a wide array of genomic variant detectors.
+
+vcflib provides methods to manipulate and interpret sequence variation as it
+can be described by VCF. It is both:
+
+    an API for parsing and operating on records of genomic variation as it can
+    be described by the VCF format
+
+    and a collection of command-line utilities for executing complex
+    manipulations on VCF files.
+
+The API itself provides a quick and extremely permissive method to read and
+write VCF files. Extensions and applications of the library provided in the
+included utilities (*.cpp) comprise the vast bulk of the library's utility for
+most users.
+
+WWW: https://github.com/vcflib/vcflib
diff --git a/biology/vcflib/pkg-plist b/biology/vcflib/pkg-plist
new file mode 100644
index 000000000000..944045d78581
--- /dev/null
+++ b/biology/vcflib/pkg-plist
@@ -0,0 +1,104 @@
+bin/LD
+bin/abba-baba
+bin/bFst
+bin/dumpContigsFromHeader
+bin/genotypeSummary
+bin/hapLrt
+bin/iHS
+bin/meltEHH
+bin/normalize-iHS
+bin/pFst
+bin/pVst
+bin/permuteGPAT++
+bin/permuteSmooth
+bin/plotHaps
+bin/popStats
+bin/segmentFst
+bin/segmentIhs
+bin/sequenceDiversity
+bin/smoother
+bin/vcf2dag
+bin/vcf2fasta
+bin/vcf2tsv
+bin/vcfaddinfo
+bin/vcfafpath
+bin/vcfallelicprimitives
+bin/vcfaltcount
+bin/vcfannotate
+bin/vcfannotategenotypes
+bin/vcfbreakmulti
+bin/vcfcat
+bin/vcfcheck
+bin/vcfclassify
+bin/vcfcleancomplex
+bin/vcfcombine
+bin/vcfcommonsamples
+bin/vcfcountalleles
+bin/vcfcreatemulti
+bin/vcfdistance
+bin/vcfecho
+bin/vcfentropy
+bin/vcfevenregions
+bin/vcffilter
+bin/vcffixup
+bin/vcfflatten
+bin/vcfgeno2alleles
+bin/vcfgeno2haplo
+bin/vcfgenosamplenames
+bin/vcfgenosummarize
+bin/vcfgenotypecompare
+bin/vcfgenotypes
+bin/vcfglbound
+bin/vcfglxgt
+bin/vcfhetcount
+bin/vcfhethomratio
+bin/vcfindex
+bin/vcfinfo2qual
+bin/vcfinfosummarize
+bin/vcfintersect
+bin/vcfkeepgeno
+bin/vcfkeepinfo
+bin/vcfkeepsamples
+bin/vcfleftalign
+bin/vcflength
+bin/vcfnormalizesvs
+bin/vcfnull2ref
+bin/vcfnumalt
+bin/vcfoverlay
+bin/vcfparsealts
+bin/vcfprimers
+bin/vcfqual2info
+bin/vcfrandom
+bin/vcfrandomsample
+bin/vcfremap
+bin/vcfremoveaberrantgenotypes
+bin/vcfremovesamples
+bin/vcfroc
+bin/vcfsample2info
+bin/vcfsamplediff
+bin/vcfsamplenames
+bin/vcfsitesummarize
+bin/vcfstats
+bin/vcfstreamsort
+bin/vcfuniq
+bin/vcfuniqalleles
+bin/vcfunphase
+bin/wcFst
+include/vcflib/BedReader.h
+include/vcflib/Variant.h
+include/vcflib/cdflib.hpp
+include/vcflib/convert.h
+include/vcflib/gpatInfo.hpp
+include/vcflib/join.h
+include/vcflib/mt19937ar.h
+include/vcflib/pdflib.hpp
+include/vcflib/rnglib.hpp
+include/vcflib/split.h
+include/vcflib/ssw.hpp
+include/vcflib/ssw_cpp.hpp
+include/vcflib/var.hpp
+include/vcflib/vec128int.h
+include/vcflib/veclib_types.h
+lib/libvcflib.a
+lib/libvcflib.so
+lib/libvcflib.so.1