Having each read go individually into the sqlite3 db is taking epically long. If we could group identical placements it would make things much more reasonable.

now made irrelevant by using a real phyloxml implementation

That is, if a sequence name appears in both the reference and query alignments.

There are a couple last things that need to be done before this merge should be done.

• clusterviz, unifrac, and parts of the old placeviz utility don't have new equivalents, and the former two are still lingering in source control.
• There should be some tidying-up all around; this has removed a lot of old code, and there's still some more stuff that can be removed. I'll see if I can find a tool that will spot dead code in ocaml.
• The command group names in guppy.ml should probably be changed or rearranged as they're still just placeholders.
• There's documentation that needs to be updated, like the readme.
• Should all references to Mokaphy_{common,base} be refactored/removed?

For non-normalized KR calculation.

they should be identical!

should "placerun" be exposed to outside?

for erick...

current name will confuse the OTU krew.

we often run of_pre-- this would seem to be slow.

All there is is https://github.com/matsen/pplacer/wiki/place_file

EDPL trees used to be available, and now they are not. They are useful.

please!

Right now, it's even keeping trailing whitespace.

Currently there's not really any formal way of tracking what changes have been done per release. We should change that!

right now,

xphyloxml is what we want to move towards
phyloxml is the old version, which assumes the bark maps have write_xml methods
myxml is my (deprecated) xml writing stuff
[the various bark files will have to be changed as well to make pxtrees]
placeviz uses them!

I have a feature request: I'm having a hard time running placeutil from wrapper scripts because of the way that output file paths are generated. Here's an example - let's say I have a placefile topdir/subdir/qalign.place, and I do this:

% cd topdir
% placeutil --distmat subdir/qalign.place

The result is this:

% ls . subdir
.:
qalign.distmat  qalign.place  subdir/

subdir:
qalign.place

A more intuitive (and easier to manage behavior, IMO) would be to write the output to subdir/qalign.distmat (and without making a copy of qalign.place in topdir). One problem is that the calling script has to do extra work to move the output file around if you don't want it in topdir; another problem is the potential for collisions of you have something like

topdir/subdir1/qalign.place  topdir/subdir2/qalign.place

and want to run --distmat on both. I'd propose either making the default behavior as described above or providing a --samedir argument to do the same thing. Perhaps even better would be to have an argument to explicitly name the output file (maybe by allowing --distmat or any other option that results in a single output file to take an argument like --distmat=subdir/whatever.distmat). Possible?

Thanks!
Noah

does the currying in make_dist_fun slow things down?

Superseded by Csv module.

once it's actually accessible from the command line...

Can we fix this, or at least give a reasonable error message?

There is some overlap, but how much?

that reports the total amount of time

• building the reference tree
• ranking the edges (totaled across all query seqs)
• playing ball (ditto)

The binaries do not run on computers without GSL.

eg in classcompare/pipeline/example:

guppy classify --csv -c vaginal_16s.refpkg refs250bp.json
grep name refs250bp_merged.class.csv | head
1:name,desired_rank,rank,tax_id,likelihood,origin
25:name,desired_rank,rank,tax_id,likelihood,origin
68:name,desired_rank,rank,tax_id,likelihood,origin
90:name,desired_rank,rank,tax_id,likelihood,origin
112:name,desired_rank,rank,tax_id,likelihood,origin
134:name,desired_rank,rank,tax_id,likelihood,origin
156:name,desired_rank,rank,tax_id,likelihood,origin
178:name,desired_rank,rank,tax_id,likelihood,origin
200:name,desired_rank,rank,tax_id,likelihood,origin
222:name,desired_rank,rank,tax_id,likelihood,origin

So that we don't divide by tree length.

I don't know if the R JSON parser generates nonstandard output, but here's the error I see when I read in a json-format file and write it back out using R:

% guppy tog -o foo merged.json
% echo "library(rjson); write(toJSON(fromJSON(file='merged.json')), file=file('merged_r.json','w'))" | R --slave
% guppy tog -o foo merged_r.json
Fatal error: exception Jsontype.Invalid_format("expected array, got string")

reinstate the -B option

• Name sucks
• Some camelcase and some underscore
• Needs reorganization
  • Perhaps some submodule action?
• Use unsafe_get when appropriate

We have alignments going to lists and to arrays, and no uniform interface.

If we want to layer these list and array functions on top a generator-like interface, that's fine, but all-at-once would be fine for now. Perhaps some mli files once it settles?

and move into a classif field

All that is really needed is to run tests/data/heat/run.sh.

This seems perhaps like a buildbot thing, as it is most easily run directly from the command line.

http://en.wikipedia.org/wiki/Power_iteration

so that we get the top several eigenvectors.

This is pretty cryptic.

Best would be to look for the version number first. If that's not found, then conclude it's not a place file and exit gracefully.

Here is the old code

let filter_prefix = match !out_prefix with
    | "" ->
      (Placerun.get_name pr)^"."^(Filename.basename (!tax_exclude_fname))
    | s -> s
    in
    let removes =
      Tax_id.TaxIdSetFuns.of_list
        (List.map Tax_id.of_string
          (File_parsing.string_list_of_file (!tax_exclude_fname)))
    in
    Placerun.multifilter
      [filter_prefix,
        (fun pq ->
          not (Tax_id.TaxIdSet.mem
                (Placement.contain_classif
                  (Pquery.best_place criterion pq))
                removes))]
      pr

but it would be nice to have inclusion and an exclusion filters.

just a little refactoring...

I'm getting a warning message from R's JSON parser when I read pplacer-generated json files:

% pwd
/home/nhoffman/working/microbiome-demo
% echo "library(rjson); jout <- fromJSON(file='p4z1r36.json')" | R --slave
Warning message:
In readLines(file) : incomplete final line found on 'p4z1r36.json'

It seems to be made happy by adding a newline to the end of the file:

% echo "" >> p4z1r36.json

Right now they only use Weighted.

but it says that it is in guppy documentation. The issue is:

:glob:

generated_rst/*

how can we exclude something?

A user would like them back: "I would like to create nexus or newick trees for other analyses. Can these be exported from pplacer?"

I think we just need to check and see if no anonymous arguments are given.

if not present, guppy should just use the common tree between the placeruns.

should be an option to perform Bonferroni

Running

guppy merge -o merged.json p1.json p2.json ... pn.json

should make a place file with the placements from p1 ... pn. There was something like this in placeutil.

There should also be a split command like in http://matsen.fhcrc.org/pplacer/manual.html#placeutil

[sometimes we don't want it to be required]