knime-mpicbg / knime-scripting Goto Github PK

used REXPFactor

executing the node leads to a terminal window with such a message:

/var/folders/aT/aT-JBgnUFAa41sIiC1a8DE+++TQ/-Tmp-/analyze6790496198141459805.py ; exit;
-bash: /var/folders/aT/aT-JBgnUFAa41sIiC1a8DE+++TQ/-Tmp-/analyze6790496198141459805.py: python: bad interpreter: No such file or directory

kIn should be a data frame object of the pandas library as these objects are much better for data manipulation

If a R script (or template) has a syntax or evaluation error, the message is always the same.
It would be helpful to wrap the script in a try catch clause before sending it to Rserve and get the variabel storing the error in case of a failure.

to would be nice to make selection of flow variables available in RGG (drop down box, ...) so that one doesnt have to edit the R script.

linux is reported as unsupported platform.
is there any reason, why?

The density is plotted as probability between 0 and 1, whereas the histogram shows counts. Often the density is not visible for this reason. A second axis would make it more useful.

File will be added to the list, though the dialog complains that the file is either empty or not accessible at all.
Should not appear in the list!

pyhton nodes fail to transfer tables with missing values for integer columns. for string columns 'nan' is returned instead of a missing value. double columns are fine.

If an rgg-template does contain the domain placeholder $$$DOMAIN('xyz')$$$ more than once, it fails to properly replace them with values and results in strange rgg-GUIs (if valid at all).

Error message:

ERROR PythonSnippetNodeModel Traceback (most recent call last):
ERROR PythonSnippetNodeModel File "/var/folders/aT/aT-JBgnUFAa41sIiC1a8DE+++TQ/-Tmp-/analyze2520056644575458498.py", line 214, in
ERROR PythonSnippetNodeModel kIn = read_csv(r"/var/folders/aT/aT-JBgnUFAa41sIiC1a8DE+++TQ/-Tmp-/knime2python2543081995884595474.csv", True)
ERROR PythonSnippetNodeModel File "/var/folders/aT/aT-JBgnUFAa41sIiC1a8DE+++TQ/-Tmp-/analyze2520056644575458498.py", line 143, in read_csv
ERROR PythonSnippetNodeModel types = get_column_types(csv_filename)
ERROR PythonSnippetNodeModel File "/var/folders/aT/aT-JBgnUFAa41sIiC1a8DE+++TQ/-Tmp-/analyze2520056644575458498.py", line 68, in get_column_types
ERROR PythonSnippetNodeModel for row in csv_reader:
ERROR PythonSnippetNodeModel _csv.Error: iterator should return strings, not bytes (did you open the file in text mode?)
ERROR Python Snippet Execute failed: No python output table found, check script output

Reason:
csv.reader(open(csv_filename, 'rb'), delimiter=',', quotechar='"')

is correct for 2.x but not for 3.x

It would be nice to have a checkbox to automatically use the local r-server. Unchecking it would then use the one in the field.

If generally the localhost is 127.0.0.1 I would suggest to use this IP address instead of the word localhost. I recently had the problem that it only worked with the IP address but not with the word localhost.

(martin)

knime.flow.in
knime.in
knime.out
knime.loaded.libraries
knime.colors

It would be nice to build up the template tree by template file sources first. At least as a view option.
And the template description should also show the source.

Processes from different nodes should be queued instead.

So the following line needs to be modified accordingly
https://github.com/knime-mpicbg/knime-scripting/blob/development_2.9/matlab4knime/src/de/mpicbg/tds/knime/scripting/matlab/AbstractMatlabScriptingNodeModel.java#L59

check if this is possible

In contrast to the MatlabPlotNodeModel and RPlotNodeModel, the PythonPlotNodeModel does not implement saveInternals and loadInternals to de/serialize the generated plots. So plots are lost after closing and reopening a workflow.

If an RGG-script has a parser error it just tells that the document is not a valid RGG which is not very helpful.

This might be an rgg-related-issue. needs closer investigation.

exclude node dependent variables (knime.script.in)

The fix column names node throws a syntax error when using strict mode and the input table contains a column with '
for example "pKi'max"

Requested by George Papadatos (Lilly UK):

... another thing that can be
sometime annoying: the python snippet node throws an error when the input
KNIME table contains cell types other than the usual string, int or double,
such as molecule or bitvector types. I guess this is because in the python
code there is a function that tries to figure out the column type. Would it
be possible to work around this problem?

from KNIME forum (Andrew):

Since upgrading to 2.10 on Friday, I am not able to edit templates. Clicking 'Edit Templates' opens the editor window, but it's read-only. I'm running Mavericks.

the node pushes an object called 'rOut' to R instead of 'kIn'

The panel which shows the RGG-GUI does not use full width/height of the window.

The KNIME image processing plugin has a "bitmask" cell type (result of object segmentation). if for instance a Rplot tries to send this data to the server it's getting stuck. Unfortunatly the server does not respond afterwards anymore. The only work around I know about until now is to restart the server.

Currently it's "plot(kIn)" which is not even valid python code. So having something like an empty coordinate system would be much better to get started.

Up to now, the naming scheme from KNIMEs old R-nodes has been supported.
This makes source code rather complicated and as these KNIME nodes are deprecated anyway, we can stop supporting them.

It would be nice to have a check box or a tab+button or something similar in a R plot node and R snippet node to use Open in R functionality from these nodes directly. Then the configured R code could be pushed to clipboard or even to a new R document where it is ready to be executed and debugged.

This would make "search Open in R node", "drag to workflow", "connect", "execute", "open R node", "unlink form template", "copy", "new R documant" ,"paste" much faster for debugging.

(Martin)

Original user report (from Oliver Gathman from Cenix Bioscience):
Mir ist zudem aufgefallen, daß wenn zwei template files die gleiche template category verwenden die category aus dem zweiten template file die aus dem ersten template file überschreibt (ich hatte versucht, in einem eigenen template file definierte snippets der in Ihrem default template file definierten category "screenmining" hinzuzufügen). Kein Drama, natürlich, aber vielleicht gut zu wissen :-) .

To reproduce the problem

1. Create two template definition files and change the knime-prefs to use both as r-template resources
2. Create a template in each file with the same category but a different knime
3. Open the node, switch to template repository --> ERROR: Just one template appears.

The R density histogram fails with too few elements. It would be more appropriate to produce a plot with a mention that too few data points are given.

Currently row IDs are lost when using R-scripting nodes. By simply using them as row.names of kIn they would be preserved and might be helpful when processing data with R.

The node fails to push them. The problems occurs when it tries to assign a variable to the REXP that contains the table.

this would allow to use the KNIME colour manager to assign color to subsets. Then in the future the plot templates could provide the option to use the KNIME color model.

ps:
Maybe this would be possible with shapes as well but then a dictonary has to be implemented in the background that turns the KNIME shape model values (e.g. "open circle" into R shape pch values "1")

It raises this error message:
IOError: [Errno 22] invalid mode ('wb') or filename: 'C:.....\Local\Temp\pyplot5587704929876224604.png

if the ouput path is not valid (e.g. non existing directory), the node fails to execute and throws a null pointer exception
see workflow: 2012-03-02_PlotWithWrongPath-NPE

I noticed that the description of template gets additional line breaks in the Script Editor after edition the template using Edit Template. Each edit/apply(ok) results in an additional line break, e.g. like this:

1. Select numerical variables for ROC curves (parameter).

"here are many line breaks"

2. Template expects a column named "treatment" as category column. Enter the name of the negative control class as plain text. Select several classes as positive controls (cases) that will be plotted against the single negative control class (controls).

I noticed that also copy/paste line breaks in the editior then make these additional line breaks.

Currently seen in tds-cruncher-1 (windows/OS-specific problem?)

scale .png or make scrollable

1. check "Don't Overwrite" option in the plot-node configuration
2. Execute -> fine, file is written
   Now:
3. Execute -> error, file exists, no plot view created

This is fine, but there should be an option to disable the output file without removing the file path.
The idea is to create another checkbox to enable or disable writing the plot to file.

From forum:

Dear Knime community,

Since I have updated the Python Snippet node, it failes even with:

kIn=pyOut

It returns this error:

ERROR PythonSnippetNodeModel File "/tmp/analyze8046917519877201488.py", line 136
ERROR PythonSnippetNodeModel d = {k: d[k].values() for k in d} # convert to dict of lists (as used by the python snippet)
ERROR PythonSnippetNodeModel ^
ERROR PythonSnippetNodeModel SyntaxError: invalid syntax
ERROR Python Snippet Execute failed: No python output table found, check script output

The Knime Python Scripting extension version is 2.0.0.201410070829

A StringCell in KNIME with e.g. '1E34' is interpreted as scientific notation by pandas instead of a string.

If an https file is given in the preferences, a new node will not be configurable anymore

ERROR Node Error while creating node dialog for 'R Plot': R-templates must start with a line with the scheme # name: blabla

If Rserve is not running (tds-rserver-2 or localhost) this error is thrown by the R plot node and R snippet node:

ERROR R Plot Execute failed: Could not connect to R. Probably, the R-Server is not running.
Here's what you need to do:

1. Check what R-server-host is configured in your Knime preferences.
2. If your host is set to be 'localhost' start R and run the following command
   library(Rserve); Rserve(args = "--vanilla")

That's fine, but in case of the Open in R node this error is thrown?
ERROR Open in R Execute failed: spawning of R-process failed

Shouldn't it be the same? ...and wouldnt it be nice to have Open in R be executable without any server running???? It is just local anyway!!?

(martin)

For convenience the KNIME table is currently parsed into a Statistical Arrays in Matlab. This requires the Statistics Toolbox!
Even though this is a fairly standard toolbox that most Matlab users choose to purchase, I would be nice to reactivate the following code
https://github.com/knime-mpicbg/knime-scripting/blob/development_2.9/matlab4knime/src/de/mpicbg/tds/knime/scripting/matlab/MatlabSnippetNodeModel.java#L63-71

And integrate the option into the configuration dialog