R error: "nlabels not found" in version 1.1.9

[EricR86]

Original report (archived issue) by Eric Roberts (Bitbucket: ericr86, GitHub: ericr86).

---

Original issue 24 created by ericr86 on 2012-01-10T16:33:36.000Z:

Running segtools-gmtk-parameters, version 1.1.9, results in the error : object 'nlabels' not found.

The error is in the definition of the function save.track.stats in track_statistics.R, line 631.

Restoring the definition from segtools version 1.1.7, corrects the issue, as nlabels and ntracks are defined within the function.

[EricR86]

Original comment by Eric Roberts (Bitbucket: ericr86, GitHub: ericr86).

---

Comment #1 originally posted by ericr86 on 2014-03-30T21:25:15.000Z:

Is there a diff for this somewhere?

[EricR86]

Original comment by Eric Roberts (Bitbucket: ericr86, GitHub: ericr86).

---

Comment #2 originally posted by ericr86 on 2014-03-31T18:33:49.000Z:

I just uploaded the Segtools repo to here:

!/hoffmanlab/segtools

Sadly, version tags did not seem to be included in the conversion for SVN (they are for Segway). You can find the Segtools 1.1.7 source linked from

https://pypi.python.org/pypi/segtools/1.1.7

[EricR86]

Original comment by Eric Roberts (Bitbucket: ericr86, GitHub: ericr86).

---

Tags were not preserved evidently on the transition from SVN to mercurial.

The last revision that I can find where track_statistics.R matches is with revision 176:53927fbc27bda83c0ae23cbba12b6f9da692e9a5. There are 17 changes to track_statistics.R between then and now (the last being on 272:cf7c6f62cf1f6ce6d7d7a3e60e702c24575ec345).

The error comes from revision 223:66a8dde3a471b4d36bddeb30b00ec1383c7d0933

$ hg log -r 223
changeset:   223:66a8dde3a471b4d36bddeb30b00ec1383c7d0933
user:        Orion Buske <buske@cs.toronto.edu>
date:        Thu Oct 13 01:30:02 2011 +0000
files:       segtools/R/track_statistics.R
description:
Made width specifiable in in track stats plotting

$ hg diff -c 223 segtools/R/track_statistics.R
diff -r 3258d3bcba65110d06ba6d2b9e040d13e9492506 -r 66a8dde3a471b4d36bddeb30b00ec1383c7d0933 segtools/R/track_statistics.R
--- a/segtools/R/track_statistics.R     Wed Oct 12 17:17:27 2011 +0000
+++ b/segtools/R/track_statistics.R     Thu Oct 13 01:30:02 2011 +0000
@@ -567,7 +567,7 @@
     col.ord <- match(label_order, colnames(means))
     stopifnot(col.ord > 0, length(col.ord) == ncol(means))
   }
-  par(oma = c(1, 1, 1, 1))  # Add a margin
+  #par(oma = c(1, 1, 1, 1))  # Add a margin

   sds.ordered = NULL
   if (! is.null(sds)) {
@@ -627,6 +627,10 @@
                              symmetric = FALSE,
                              clobber = FALSE,
                              square.size = 15,  # px
+                             height = 200 + square.size * ntracks,
+                             width = 450 + square.size * nlabels,
+                             height.pdf = height / 72,
+                             width.pdf = width / 72,
                              ...) {
   mnemonics <- read.mnemonics(mnemonic_file)
   translations <- read.mnemonics(translation_file)
@@ -637,13 +641,13 @@

   ntracks <- nlevels(res$stats$trackname)
   nlabels <- nlevels(res$stats$label)
-  height <- 200 + square.size * ntracks
-  width <- 450 + square.size * nlabels
-
+
   save.images(dirpath, namebase,
               levelplot.track.stats(res, symmetric = symmetric, ...),
               height = height,
               width = width,
+              height.pdf = height.pdf,
+              width.pdf = width.pdf,
               clobber = clobber)
 }

Essentially the defined width/height parameters were copy pasted into the parameters probably not realizing that using nlabels in the default parameters was a bad idea. There's a good chance it was never caught if the function was always called with non-default width and height parameters.

This was never fixed in later histories (and then accidentally reverted).

It does raise the question as to what the default width/height parameters should be.

[EricR86]

Original comment by Michael Hoffman (Bitbucket: hoffman, GitHub: michaelmhoffman).

---

Can you fix it with something like height = NULL, width = NULL, height.pdf = NULL, width.pdf = NULL in the argument list?

Then after ntracks and nlabels are defined you can can do something like if (is.null(height)) height <- 200 + square.size * ntracks? (Better to have that 200 as another argument if so.)

[EricR86]

Original comment by Eric Roberts (Bitbucket: ericr86, GitHub: ericr86).

---

Can you clarify why you would like the 200 as a default for a function argument? If you're referring to the default that's already there I don't see why it couldn't be used in the default when height = NULL.

[EricR86]

Original comment by Michael Hoffman (Bitbucket: hoffman, GitHub: michaelmhoffman).

---

As a general rule, we shouldn't have magic numbers like that in the body of a function. A constant at the top of the file is fine.

[EricR86]

Original comment by Eric Roberts (Bitbucket: ericr86, GitHub: ericr86).

---

• changed state from "new" to "resolved"

Fixed by commit d46049a6e1f18a462cdac8eb5505046f50e41324