Ok, some lose ends remain, but this involves a lot of bug fixes and speedups, so pushing this through now, and will pick up the remnants shortly.
Changes to Phyx
Repo-wide
the biggest thing you'll notice is the removal of using namespace std; throughout (including header files :/). this is to avoid potential namespace conflicts. instead, things have been updated to use explicit namespace qualifications. so for example, use std::vector<std::string> and std::cout instead of vector<string> and cout. this involves more typing, but is safer. and honestly you get used to it.
if this is too awkward, use an explicit (and limited) using-declaration like using std::vector; and using std::cout; to allow easy typing of vector<int> and cout, but limit the use of these and their scope as much as possible
more extensive alignment reading capabilities, and better error-checking. this includes annoying variants of the phylip format, and support for morphological data sets (i.e., with arbitrary chaaracter state labels). for programs that can work on streams (i.e., one sequence at a time) the reader has been updated. for programs that require the entire alignment to be read in before a task can be performed, the function std::vector<Sequence> ingest_alignment (std::istream* pios, std::string& alphaName); (in seq_reader.h) can be used. here, alphaName is the phyx alphabet type (DNA, AA, BINARY, MULTI), passed by reference, as this is often important information for downstream procedures.
when writing a new program, use the python script generate_manpages.py to generate its manpage. this uses help2man, which is easy to install. as long as your px* -h output is appropriate (see existing ones) then decent manpages will be geenrated
the phyx Makefile will (un)install manpages with programs
no one will ever use this feature, but it is there!
all warning/error messages are written to std::cerr instead of std::cout
"orphan files" (i.e., those not currently used in the Makefile in any way) are moved to the directory /src/orphan_files. some of these are deprecated, others are only half implemented (and perhaps forgotten). please resurrect as applicable
all programs now have a -C --citation option. currently these all report the phyx Bioinformatics paper. additional citations should be added for indiviual methods as applicable.
given the extensive changes, and in preparation for a new release, i've updated the version of the individual programs to v1.1
this is an issue that appeared in a paper for the R package castor
the issue involves the extensive memory management overhead involved with deleting nodes. pxrmt now will use the pxtrt tracing procedure is the number of nodes to remove are too large
the castor example now takes 5 seconds (instead of more than 3 hours; the author actually killed the job at this point)
pxrms (and pxrmt) now uses regex for removing taxa (feature request)
New programs
pxmono: test the monophyletic status of a list of names in a tree (or distribution)
pxcomp: sequence alignment compositional heterogeneity test (chi-square)
pxtgen: generate exhaustive (rooted or unrooted) tree topologies for n taxa (limited to 10)
for hypothesis testing
Miscellaneous (no particular order)
RAII timer added with timer.h for quick 'n easy code timing
to use this, do something like:
include "timer.h"
...
void foo (some args) {
// some initializing code here
{
Timer tm;
// do something here to be timed
}
// do something else
}
- `pxupgma` was completely rewritten. somewhere along the line, it stopped producing the correct result (using PAUP* as a check)
- `pslssq` now handles all the supported datatypes (DNA, AA, BINARY, MULTI) rather than just the first 2
- `pxboot` overhauled
- `pxclsq` refactored
- added the ability to peek at line(s) in a stream to make decisions for downstream procedures without losing position in the stream. found in `utils.h`:
- `pxaa2cdn` was overhauled (more efficient, and better error-checking)
- more consistent member variable names and formatting
- remedy various `-Wall` compiler warnings
- tree cloner in `tree.h`. important if we want to do something with a tree _and_ keep the original
- `pxs2nex` write the correct Nexus datatype and alphabet
- a bunch of new example files in `/src/TEST`. all the crazy phylip formats, interleaved Nexus files, binary and morphology alignments, etc.
- `pxtrt` overhaul
- about 40K other minor changes
Ok, some lose ends remain, but this involves a lot of bug fixes and speedups, so pushing this through now, and will pick up the remnants shortly.
Changes to Phyx
Repo-wide
using namespace std;throughout (including header files :/). this is to avoid potential namespace conflicts. instead, things have been updated to use explicit namespace qualifications. so for example, usestd::vector<std::string>andstd::coutinstead ofvector<string>andcout. this involves more typing, but is safer. and honestly you get used to it.using std::vector;andusing std::cout;to allow easy typing ofvector<int>andcout, but limit the use of these and their scope as much as possiblestd::vector<Sequence> ingest_alignment (std::istream* pios, std::string& alphaName);(inseq_reader.h) can be used. here,alphaNameis the phyx alphabet type (DNA, AA, BINARY, MULTI), passed by reference, as this is often important information for downstream procedures.#includes throughoutgenerate_manpages.pyto generate its manpage. this useshelp2man, which is easy to install. as long as yourpx* -houtput is appropriate (see existing ones) then decent manpages will be geenratedMakefilewill (un)install manpages with programsstd::cerrinstead ofstd::coutMakefilein any way) are moved to the directory/src/orphan_files. some of these are deprecated, others are only half implemented (and perhaps forgotten). please resurrect as applicable-C --citationoption. currently these all report the phyx Bioinformatics paper. additional citations should be added for indiviual methods as applicable.v1.1Issues addressed
pxclsq(feature request)pxtrtleaving a root edge (bug)pxrmtspeed (enhancement)castorpxrmtnow will use thepxtrttracing procedure is the number of nodes to remove are too largecastorexample now takes 5 seconds (instead of more than 3 hours; the author actually killed the job at this point)pxrms(andpxrmt) now uses regex for removing taxa (feature request)New programs
Miscellaneous (no particular order)
timer.hfor quick 'n easy code timingvoid foo (some args) { // some initializing code here { Timer tm; // do something here to be timed } // do something else }
std::string peek_line (std::istream& pios); std::vector peek_lines (std::istream& pios, const int& n);