Hi, A screenshot isn't very helpful. You have to find the logs and look there inside for the root cause. Failing at start means you should probably look into all the components which are being started to find the one (or many) being the problem.

You can't have a mixed datatype in the same column: if 'a' is a number, 'N/A' should be a number (which is impossible). If you really want 'N/A' then every single number in that column should also be a text, which is maybe not what you want... CAST() can be used to convert any datatype.

Can you look what the physical query looks like in the logs? Depending on the database used DATE is maybe not without time component, therefore even a CAST isn't changing anything at the real values.

I don't have a column with a time component handy, but you could try using CASE(<your column> as DATE). So in your case the filter would become CAST(<column 1> as DATE) = CAST(<column 2> as DATE)

Are those dates really just dates? You seeing '22-Dec-20' doesn't mean the real value isn't 2020-12-22 23:53:47.845 And if you have a time component in your dates, there are lot of chances they will not match the comparison as they will probably all have different times.

Hi, Is that an analysis or a report (BI Publisher)? Assuming you mean analysis (by the name of your columns), you take your first column and add a filter. There you check the "Convert this filter to SQL" checkbox. In the window you have, you can enter the SQL condition you want to use. Just write your column1 = column2…

Hi, You are already getting answers in the MOSC thread you open. Keeping it short: yes you can have as many installs as you want in a single server as long as they don't overlap on disk and that running environments don't conflicts in ports (so same ports are fine if the 2 or more instances aren't running at the same…

The ideal solution is to have a proper time dimension containing your own business rules for dates, because start and end date of a week is a something which must follow your own rules and is going to be different for other companies. If you want to do that in OBIEE itself you can calculate things by using the date…

You maybe misread what Joel was saying: what does the physical SQL looks like? It must be a valid query as you get something back, but what is it like? Because if the physical query is correct the result can't be wrong (ok, it can if you have selection steps or other things happening at the BI Server level, but you didn't…

The 10g piece of code was hand-made by somebody who shared it. For 11g somebody did it for sure, but they didn't share it (or maybe they did, no idea where).

And if you change the aggregation to count distinct in the front-end? Or create a new calculated column in the analysis?

Set your measure as a COUNT DISTINCT aggregation in the RPD and job done, no? Even if you display it multiple times because that's the reality (the same number exist in multiple dates or weeks), the totals for higher levels will do a count distinct and will work just well.

It has been asked and it has been answered, at least 4-5 times (these are those I answered myself :P ). Just like you found, there isn't something out of the box for that. You have to write you own piece of CSS & JS for that. A hand-made piece of code you have to write and maintain for your needs.

Use a narrative view to display a column of your analysis in which you have the logic. Or the same narrative to have a piece of javascript to display or hide a static piece of text. There are many ways to get there....

Do you mind writing all the pieces you forgot in your thread? Like product names, context and what you try to do? "Oracle 11" isn't really a product as Oracle has about 100+ products which have a version 11...

As Christian said I was staying away of this because OBIEE in French just doesn't work (not technically): all the messages and errors have little to no meaning in French as things have been translated fairly random (and I do speak French fluently!). That error is a fairly common one, mainly when adding a new LTS and…

It's simple logic, if you don't store the right format in your database, you just have to build it by yourself. So with some basic logic you could multiply YEAR(date) by 100 and you sum it with MONTH(date) and you can finally turn the resulting number into a char if that's the format you are after.

If you look at the available functions in OBIEE there is a whole set about dates, including extracting components from a date. Just look in the UI and you will see the list of functions available.

In the RPD you should set the sort order column for your text month logical column. Pick a column which can be sorted correctly for months (like the month number or something like that) and use that one as sort column for the text versions of months.

I used "LIKE pattern" in prompts and created a presentation variable and used this as a filter with the same "LIKE Pattern" criteria in the filter.It is not working. And why it isn't working? What does the generated SQL query looks like? What should be different in that query to make it work? You are maybe just 1 step away…