Errors (x2) whilst trying to export MDS analysis to CSV in certain circumstances

[wydnas]

The error message as detailed as possible
ERROR 1:
Could not open the file.
KH Coder will exit now.

• C:/khcoder3.17k/config/R-bridge/khc2_word_mds_1_save.csv

/<C:\khcoder3.17k\kh_coder.exe>kh_r_plot/mds.pm line 55

Related Error 2:
No error message but csv output from MDS graph is incorrect as it contains the data from the last succesful previous export and not the current export once a word limit threshold is crossed (see below for commentary)

Console:
ERROR 1:
Encoding of this Console: cp850
Encoding of this file system: cp850
No type library matching "Microsoft Word" found at /<C:\khcoder3.17k\kh_coder.exe>kh_docx.pm line 94
Win32::OLE(0.1709): GetOleTypeLibObject() Not a Win32::OLE::TypeLib object at /<C:\khcoder3.17k\kh_coder.exe>Win32/OLE/Const.pm line 49.
M$ Word not found...
Locale: cp1252
This is KH Coder 3.Alpha.17k on MSWin32.
CWD: C:/khcoder3.17k
Available Physical Memory: 2047MB
Checking MySQL connection...
Starting MySQL...
coder.ini not changed. skip saving...
win.ini not changed. skip saving...
2020-02-28 11:05:35 0 [Note] bin\mysqld (mysqld 5.6.39-log) starting as process 11568 ...
R Version: 3.1, x86_64
Using un-threaded functions...
Monitors: 0, 1366, 0, 768
new window: 396, 139
new window: 112, 135
Connected to MySQL 5.6, khc2.
Checking icode (en)... cp1252 or iso-8859-1

Server cmd: java -showversion -mx300m -cp "C:/khcoder3.17k/dep/stanford-postagger/stanford-postagger.jar" edu.stanford.nlp.tagger.maxent.MaxentTaggerServer -outputFormat xml -outputFormatOptions lemmatize -port 32020 -model "C:/khcoder3.17k/dep/stanford-postagger/models/wsj-0-18-left3words-distsim.tagger"
Starting server, pid: 12396, Connecting.openjdk version "1.8.0_212"
OpenJDK Runtime Environment (AdoptOpenJDK)(build 1.8.0_212-b03)
OpenJDK Server VM (AdoptOpenJDK)(build 25.212-b03, mixed mode)

Loading default properties from tagger C:/khcoder3.17k/dep/stanford-postagger/models/wsj-0-18-left3words-distsim.tagger
Reading POS tagger model from C:/khcoder3.17k/dep/stanford-postagger/models/wsj-0-18-left3words-distsim.tagger ... done [1.2 sec].
. ok. Tagging... ok.
Morpho1 8 wallclock secs ( 0.49 usr + 1.38 sys = 1.86 CPU)
Single row: 348 bytes
The HEAP table will eat approx. 4MB; We have 1547MB max.
Read 0 wallclock secs ( 0.00 usr + 0.00 sys = 0.00 CPU)
Format 1 wallclock secs ( 0.06 usr + 0.00 sys = 0.06 CPU)
Strat1 1 wallclock secs ( 0.08 usr + 0.02 sys = 0.09 CPU)
Strat2 0 wallclock secs ( 0.02 usr + 0.00 sys = 0.02 CPU)
RawTXT 0 wallclock secs ( 0.03 usr + 0.00 sys = 0.03 CPU)
df: heap df: heap df 1 wallclock secs ( 0.02 usr + 0.00 sys = 0.02 CPU)
fc 0 wallclock secs ( 0.00 usr + 0.00 sys = 0.00 CPU)
Check 0 wallclock secs ( 0.01 usr + 0.00 sys = 0.01 CPU)
fix_morpho 0 wallclock secs ( 0.02 usr + 0.00 sys = 0.02 CPU)
fix_unknown 0 wallclock secs ( 0.00 usr + 0.00 sys = 0.00 CPU)
fix_headings 0 wallclock secs ( 0.00 usr + 0.00 sys = 0.00 CPU)
make_cache 0 wallclock secs ( 0.03 usr + 0.00 sys = 0.03 CPU)
clear_tmp 0 wallclock secs ( 0.02 usr + 0.00 sys = 0.02 CPU)
MySQL: FLUSH
Morpho File: C:/khcoder3.17k/config/khc2/khc2_ch.txt
RAM: 16309
new window: 60, 83
ignore:
..
0 wallclock secs ( 0.02 usr + 0.00 sys = 0.02 CPU)
Data matrix for R: 71 words x 105 (105) docs
kh_r_plot: Loading Cairo...
Loading required package: sp
Statistics::R::Bridge::pipe::read_processR, Sleep and Retry!
Checking rgeos availability: FALSE
Note: when rgeos is not available, polygon geometry computations in maptools depend on gpclib,
which has a restricted licence. It is disabled by default;
to enable gpclib, type gpclibPermit()
Loading required package: Rcpp
Loading required package: RColorBrewer
Non-function objects are not currently inserted (not traceable): .packageName
Modified functions inserted through trace(): wordlayout
Warning: package 'ggplot2' was built under R version 3.1.3
Statistics::R::Bridge::pipe::read_processR, Retry:
-------------------------[Begin]-------------------------[R]
initial value 39.215802
iter 5 value 32.128840
iter 10 value 31.676174
iter 15 value 31.496081
iter 20 value 31.303166
iter 25 value 31.238127
iter 30 value 31.199697
iter 35 value 31.187581
iter 40 value 31.185336
final value 31.184943
converged
[1] "iterations: 257"
---------------------------------------------------------[R]
done: 00:00:04
new window: 106, 40
Non-function objects are not currently inserted (not traceable): .packageName
catched:
catched:

Error 2: Not provided

To Reproduce
Steps to reproduce the behavior:

1. Load project
2. Go to Tools > Words > Multi-Dimensional Scaling
3. Run as per normal with default options
   
4. From the MDS graph screen Select Save, and choose to save to CSV format. Then either no error, error ,1 or error 2 will occur depending on circumstances.

The reported errors seem to occur only if the 'Number of Selected Words' exceeds a certain value. Originally I had thought the failure value was directly represented by the 'Number of Selected Words' but have since found that by altering the Min TF value I can get more words to be used and output to csv with success, but I haven't got it above 109 with Min TF of 5, or 84 with a min TF of 6. If I drop the Min TF to 4 then I get 144 words selected - the CSV export process appears to work without error BUT the file created under a new name appears to be a duplicate of the last succesful and correct file previously exported, which in this case contained only 109 words. The Max TF value although altering the 'Number of Selected Words' does not seems to influence in the same way as the Min TF curiously = e.g. setting the Max TF to reduce the number of selected words does not affect the ability to export if it has already failed and no other parameters are changed.

Expected behavior
Successful export of .csv file containing the correct Number of Selected Words and their assigned groupings.

What happens if you perform the same operation with the tutorial file (Anne.xls)
I can't reproduce error 1 but I can reproduce error 2. In my tests with Anne.xls I created out put with 116, 143, and 293 selected words by varying the Min TF only. The output for 116 was correct but the output for 143 and 293 words contained the results obtianed for the 116 words, for other runs below 116 words the output was correctly creted (both before and after the incorrect 143/293 attempts). So it looks like the limit is somewhere between 116 and 143.

Files to Reproduce the Error
Simmons_Geoarabia2007_FullCleanV2.txt
This file was sourced from a pdf download from the internet and has been cleaned up to remove any obviously problematic text such as diagram content text, headers and references.

Screenshots
If applicable, add screenshots to help explain your problem.

Version number of KH Coder
3.alpha 17k (and 3.alpha 17j)

OS
Windows 10 64bit

Additional context
Trying to extract word lists of results along with their group assignment. As an aside it would be really good if similar csv output (with correspondence group) could be created for the Co-Occurrence Network analysis. I am aware that 100+ words is not a normal advised amount for either MDS or Co-Occurrence Network output - but it is where I like to start before narrowing down.

[wydnas]

Using my own .txt file as supplied above, I managed to export 126 words with a min TF of 4 (but it should be noted that a different POS combination that gave 128 words also with a Min TF of 4 failed to export correctly giving error2). By altering the Min TF down to 3 on the successful 126-word export I successfully exported 180 words (187 words selected with 7 words ignored). By reducing Min TF to 2 (all other parameters the same) 328 words were selected (50 were ignored) and the CSV output was successful for 278 words. Further reducing Min TF to 1 gave 805 words selected, a lot of words were ignored (100+) and the output failed with error 2 - the output file contained the 278 words from the previous Min TF 2 output. So a bit closer, but still unable to pin down what factor triggers error 2.

Here's a screengrab that produces incorrect export to CSV (error 2)

The screengrab below uses exactly the same data but the Min TF is 2 rather than 1 and it produces a correct CSV export, 50 words are ignored.

[ko-ichi-h]

Thanks for reporting the error with detailed information. It really helps.

I think I have fixed these errors in version 3.Alpha.17L. Please download & try it out.

These error occured when there were words with same spelling and different POS like "correlative" in this case.

[ko-ichi-h]

If you want to make KH Coder recognize "correlative"s as one word, please see #96.