Sandraros/check demo regressions 2

[sandraros]

Fix #894, alternative to #902.

This is a lightweight solution, the demo programs are not changed except those ones which write sy-datum and sy-uzeit (EDIT: and demo programs which read database tables to produce Excel content → fixed data provided).

[AndreaBorgia-Abo]

Thank you for your patience with my review speed! Here are my observations:

1. not requiring an API of sorts to write a test report is arguably a pro, since users might want to adapt a test report when starting out with this library
2. we still need to account for date and time, as already noted elsewhere, but even this might not be enough: for example, in ZEXCEL_DEMO3 data coming from table SFLIGHT is different between NPL and A4H demo systems. I don't have a suggested fix here: ATDF is not supported on all versions, rewriting the tests is perhaps not worth the effort anyway since developers are the target users here... I guess we'll simply have to document it somewhere in the program itself or via a popup
3. going through the GUI probably restricts the demo checker to the Windows version: whereas the checker from the older PR would run just fine with the Java GUI, this one has the typical UNKNOWN_DP_ERROR control issue (this time in ZEXCEL_DEMO28)
4. in A4H (for this PR) all Reader tests fail because of a spurious tag: ht="0.0000000000000000E+00"
5. in NPL (for the older, as reference only) many tests fail because of this extra tag: width="9.10" (apparently coming from CREATE_XL_SHEET but I suspect that's because of the iterator changes which were made afterwards)
6. if the output directory already contains the test files, the checker cannot run; I think we could either create a temporary directory for each run or just allow overwriting the files, perhaps informing the user once

All in all, this is a better solution and you may un-draft this PR.
With regards to the older one, I think we could close it without merging and perhaps conserve the corresponding tree as food for thought, if anything else.

[sandraros]

Thanks to you! :)

1. So, closing the other solution (Sandraros/check demo regressions #902) and keeping the branch for a few days, in case...
2. I'd propose to have a hidden parameter in ZDEMO_EXCEL3 to generate hardcoded lines. I will push something.
3. Difficult to solve UNKNOWN_DP_ERROR in ZDEMO_EXCEL28/Java GUI.
4. I created Writer: useless null row height (ht="0.0000000000000000E+00") #944. I'll propose a PR quickly.
5. The iterator fix has added width="9.10" for all columns and the Excel files in the Web repository should reflect that, so I wonder if you have pressed the Refresh button to call ZDEMO_EXCEL/regenerate the demo files in the output directory.
6. Why do you say "if the output directory already contains the test files, the checker cannot run"? I don't have any issue.

[AndreaBorgia-Abo]

2. not sure if this is the only report affected but we'll soon find out :)
3. I know, but this solution is so much better that I'm willing to live with it
4. regeneration should not be needed: I wasn't jumping from one branch to the other, rather I had pulled the older PR into NPL and the newer in A4H so comparing the behaviour was a simple matter of switching window
5. because I misinterpreted the error messsage, that's why: it only happens if you leave Excel open on any of the test files, in which case it will just dump

[AndreaBorgia-Abo]

Almost there but not quite... and the test results are weird!

In A4H only:

• ZDEMO_EXCEL3
• ZDEMO_EXCEL10
• ZDEMO_EXCEL22 (also: top 3 rows are frozen, instead of just one, perhaps something was removed above)
• ZDEMO_EXCEL33

Both in NPL and in A4H:

• ZDEMO_EXCEL15, all reader tests (but that is Writer: useless null row height (ht="0.0000000000000000E+00") #944 so no worries)

This time I pulled the same branch in both systems, of course. Tomorrow I'll look into the A4H issues, unless you already have an idea.

[sandraros]

In A4H only: all these demo programs get data from tables SFLIGHT or T005T. Let me provide same solution as I did for ZDEMO_EXCEL3 (static data not from tables).

EDIT: fixed.

[AndreaBorgia-Abo]

Adding suggestion here, as already noted it's not possible to suggest changes elsewhere:

columns->get_column( 'COMPARE_XLSX_JUST_NOW' )->set_technical( ).
columns->get_column( 'COMPARE_XLSX_REFERENCE' )->set_technical( ).

[sandraros]

Adding suggestion here, as already noted it's not possible to suggest changes elsewhere

It's brand new program, all lines are new, are you sure it isn't possible to suggest changes?

[AndreaBorgia-Abo]

It's brand new program, all lines are new, are you sure it isn't possible to suggest changes?

I did not see it in the changes, perhaps because it was older. Found it now and added, thanks. Oh, another suggestion: a call transaction SMW0 to download the existing reference file, a companion function to the save button.

Anyhow, there are still some differences reported in A4H:

• ZEXCEL_DEMO3: sort order of the second sheet (Data validation) is different, but all values are present
• ZEXCEL_DEMO33: as far as I can tell the visible content is unchanged but internally some nodes still have different numbering(?)

I think we could merge this and then later understand where those differences come from.

[sandraros]

I understand, it's the column headings from SFLIGHT/T005T data elements, they may differ in each SAP version, we should fix them also.

[sandraros]

Agreed to merge now if you wish. I'll work on the fix this evening.

[AndreaBorgia-Abo]

Agreed to merge now if you wish. I'll work on the fix this evening.

Regarding the remaining differences, would you like a single ticket for both NPL and A4H or would you rather keep them distinct?

[sandraros]

Not sure what you mean. I think I'll do one PR (didn't have free time yesterday) to make NPL and A4H produce the same result. In ZDEMO_EXCEL3 I forgot to fix table data instead of SELECT * FROM SCARR, and in ZDEMO_EXCEL33 I think we need to fix the line of column headings with set_cell_value( ... ) right after calling the convert method.