search

mwang87 / SMART_NMR

Repository to help develop SMART
https://smart.ucsd.edu/
MIT License
0 stars 0 forks source link

Incorrect Input File #106

Closed beowulf-zc closed 4 years ago

beowulf-zc commented 4 years ago

Some users always forgot to make 1H data two decimals and 13C data single decimal. They have contacted me about wrong output. I would like add an error message to the test result page if their CSV format is wrong and no peaks show up on the HSQC panel. Perhaps something like "wrong input csv format, please check the decimals of your input file".

mwang87 commented 4 years ago

We don't validate this. Does this actually matter?

If it does matter, can we just truncate sigfigs?

Can you provide examples where this makes an appreciable difference?

beowulf-zc commented 4 years ago

imageThis guy first made wrong format looks like this. The result looks like this image image . Then I told him to make the input look like this. image Then everything became right.

mwang87 commented 4 years ago

Please provide the test cases in both scenarios so we can test.

beowulf-zc commented 4 years ago

tests.zip Ming, please see the file for the test data. The wrong format was the csv that was wrong. The correct input is the format that works. Please let me know if you have any questions.

mwang87 commented 4 years ago

Can confirm the wrong format creates incorrect output. What is wrong about the format?

RaphaelR87 commented 4 years ago

I just want to comment on this issue, because I think it is important for clarification.

1) the issue here is/was not the number of decimals, but I rather think it is a problem that excel might create when converting tables to .csv files that contain values with 0 decimals such as 68 instead of 68.0

When I ran the 'wrong input' with our peak list entry function the results are identical to the 'correct input' file https://smart.ucsd.edu/resultclassic?task=f5ba2b6c-6854-44f8-9333-3261699043cc

2) Independent of this issue we should still work on why SMART shows sometimes slightly different results when tables contain more than 1 decimal for 13C. To avoid/reduce that problem I updated the documentation including KyoBins suggestion for customized table export with MNova that directly converts into the desired format

mwang87 commented 4 years ago

I agree we should understand when the precision matters. To explore this can we take a spectrum and vary the precision so we can see the diference in results?