Apache OpenOffice (AOO) Bugzilla – Issue 4811
No drag and drop from ODBC- accessed databases
Last modified: 2006-05-31 14:29:06 UTC
I have already filed issue No. 4767 against ODBC- access of PostgreSQL databases. This one may be related. The Drag-and-drop mechanism described in the on-line help works perfectly from spreadsheet-based databases, from JDBC-accessed Postgresql databases, but not from the same databases accessed through ODBC. The process described in the on-line help seems to work perfectly, but nothing happens when one clicks OK at the end of the attribute selection. For the record: I get write access through ODBC, but a read-only access through JDBC; in both cases, there is no distinction between tables and views ad the beamer level, while this distinction exists in the "tables" tab of the database management tool; In consequence, while a view can be created, it cannot be saved "officially" (the saving process ends up with an unspecified error message box) or edited. See issue 4767.
Emmanuel, I kindly ask you to refine your description :) I'm pretty sure that the online help does describe a lot of DnD mechanisms, even in the database access area, there might be quite a few. Okay, I did not check this :) My point is that your current description requires the reader to guess what you are doing, and forces it to check what the online help describes. If you'd please describe what exactly you are doing, what you expect to happen, what happens though you don't it expect :), what does not happen, and so on, then this would help a lot. This, additionally, would have the advantage that the description can stand for itself - at the moment, it depends on a external resource (the online help) not changing. But before you do this, please check if issue 3045 is what you mean - I strongly suspect this from what I guessed from your description :) Thanks Frank
[ description refinement ] Uh ? I was aware of only one, namely clicking on the upper left corner of the table data window in order to slect the table data, then dragging with the left mouse button pressed to the writer window. This pops up a dialogue box in which you can select the attributes (columns)you wish to insert and the way you wish them inserted (import to table, to fields or as raw text); in order to terminate your selection and have the data inserted in your writer document, you click the OK button of the dialogue box. [ ... duplicate of 3045 ? ] Well, ... yes and no! My report of what happens is indeed close to the description given in 3045, but there is a large difference: the mechanism is functional with spreadsheet-based or JDBC-accessed PostgreSQL databases, but not with ODBC-accessed databases. Therefore, it does not look as a pure writer bug, but rather as a database- related (probably ODBC- related) problem. Since I have found several problems related to ODBC-accessed PostgreSQL databases, I wink that they may have a common cause.
Emanuel, I admit I did not look into the help, so I do not know what's actually in there. But you can drag the whole table, rows only, columns; and in all cases there exist two modes (differentiated by pressing modifier keys). For the duplicate: issue 3045 does state nothing about spreadsheets and so on, but only tells about the functionality not working with PostgreSQL-ODBC. So where's the difference? The more as other bugs (which had been flagged as duplicate of bug 3045) then describe the same as you here: _not_ working with PostgreSQL-ODBC, but with anything else. So I really like to mark this here as a duplicate .... for not finding the issue: No need to argue (TM). Searching Issuezilla is a pain, IMVHO.
you're right, I forgot to specifies that selecting only some rows does not change the problem. I was not aware of the existence of modifier keys. However, my report, in my opinion, pinpoints the problem at the ODBC-access level; there might well be other, writer-related problems, but this one is database- specific. Does this grant the existence of a separate report ? I do not know, but I would really like to see it flagged as a database issue, more specifically ODBC-related.
Emmanuel, okay, for the moment let's keep aside the component (below ...). issue 3045 says "can't do the DnD thingie with PostgreSQL-ODBC in a Writer document on Linux, but into Spreadsheets it works". issue 4811 (this one here) says "can't do the DnD thingie with PostgreSQL-ODBC in a Writer document on Linux, but with JDBC the same thing works". Is this correct? This alone would not justify a separate issue. Both descriptions do not contradict, and as long as nobody proves that both issues are different, we have to assume that they are the same (because it has a high probability, imo). Instead, the usual way is to mark as duplicate, and, if you wish, add the additional information to the original bug. So 3045 would then be "can't do the DnD thingie with PostgreSQL-ODBC in a Writer document on Linux, but into Spreadsheets or connected to the same with JDBC it works". I simply see the problem of uneccesary bugs. 3045 and 4811 very much sound like the same, so I see no reason in keeping them different, and the (ver probable!) set them to "fixed" at the same time. Now for the component: Well, "component" in general is difficult. I pretty much assume that the bug is _also_ located in the writer. In the scenario described, we act as client providing functionality for the Writer, and the fact that (nearly) the same client role works perfectly in Calc suggests that part of the problem is located in the Writer. On the other hand, there is some chance that the ODBC bridge has a problem with the PostgreSQL-ODBC driver, which is uncovered in Writer only. Wouldn't really surprise me. The reason why I originally assigned 4811 to Writer (besides that I suppose part of the problem is there) is that investigations have to _start_ there. Normally, we would need something like "Database/Writer interaction" :). But we haven't, so we have to decide for something ... Okay. I admit my argueing for the "Word processor" component may have been inspired by the technical background. From a user's perspective, it's likely that user's recognize this as "Database Access" bug instead. So I would agree to change the component field of 3045 back to "Database Access". But _then_ 4811 really is a duplicate, isn't it? :)
[ Edited transcript of my mail answer, as requested by F. S. ] >Emmanuel, > >okay, for the moment let's keep aside the component (below ...). > >issue 3045 says "can't do the DnD thingie with PostgreSQL-ODBC in a >Writer document on Linux, but into Spreadsheets it works". > >issue 4811 (this one here) says "can't do the DnD thingie with >PostgreSQL-ODBC in a Writer document on Linux, but with JDBC the same >thing works". > >Is this correct? >This alone would not justify a separate issue. Both descriptions do >not contradict, Right. > and as long as nobody proves that both issues are >different, we have to assume that they are the same (because it has a >high probability, imo). Wrong, IMNSHO. 3045 states that, with the same data source (namely an ODBC-accessed PostgreSQL database), one can drag-and-drop INTO a speadsheet, but not into a text. This tells us that writer has an issue with getting data that calc has not. Therefore, it could be a problem into either writer code or writer UI. On the other hand, 4811 tells us that one can DnD into text FROM a JDBC-accessed PostgreSQL database or FROM spreadsheets accessd through the database interface, but not FROM an ODBC-accessed PostgreSQL. Now, this tells us that something differs between ODBC access and other access(es ?).Therefore, something is wrong in data access mechanism from ODBC that is not wrong in either JDBC-access or Spreadsheet-through-database interface. Now, it might well be teh case that there is indeed a bug in writer UI or writer code that shows *only* in the case of ODBC access. But this would be sufficient to demonstrate that what data-access presents to the client application (writer, in this case) is different in the ODBC case that it is in the JDBC or Spreadsheet cases. This lack o uniformity is, in itself, a bug (or a misdesign), and it is related to data-access mechanism. Q.E.D. > Instead, the usual way is to mark as >duplicate, and, if you wish, add the additional information to the >original bug. >So 3045 would then be "can't do the DnD thingie with PostgreSQL-ODBC >in a Writer document on Linux, but into Spreadsheets or connected to >the same with JDBC it works". IMHO, no. See above. >I simply see the problem of uneccesary bugs. 3045 and 4811 very much >sound like the same, so I see no reason in keeping them different, >and the (ver probable!) set them to "fixed" at the same time. Not so probable, IMHO. See above. >Now for the component: >Well, "component" in general is difficult. I pretty much assume that >the bug is _also_ located in the writer. In the scenario described, >we act as client providing functionality for the Writer, and the fact >that (nearly) the same client role works perfectly in Calc suggests >that part of the problem is located in the Writer. You might well be right. But, IMHO, this demonstrates the existence of TWO* problems, on in data-acces mechanism, and one in Writer. >On the other hand, there is some chance that the ODBC bridge has a >problem with the PostgreSQL-ODBC driver, which is uncovered in Writer >only. Wouldn't really surprise me. Hmmm ... That would be a *third* bug, in the specifics of PostgreSQL ODBC mechanism, i. e. the ODBC driver. This, of course, is possible. However, other checks show that at least, the basic ODBC mechanisms are working well in this driver. To test this : install, for example, the R statistical software package and the RODBC packge, which provides ODBC access to R. [ Digression, but see below ... ] This is, BTW, the way I have been able to pinpoint another data-access problem to PostgreSQL databases through ODBC (see issue 4767 : no distinction between table and views in database Beamer). In R with RODBC, one can launch arbitrary SQL queries and some selected ODBC operations ; one of them is "List tables", which returns (as specified) a table of all tabes and views in the database. One of the columns of this table is a field specifying if the table is indeed a table, a system table or a view. One can easily demonstrate that this information can be correctly used by OpenOfice : in the "Tables" tab of the "Mmanage databases" tool, on can see that tables are marked with a small "table" icon, whereas views are marked with an icon bearing a broad arrow through them. By the way, the data interface of StarOffice 5.2 did not have this problem. [ End of digression ] The previous example shows that some changes made in the ODBC-specific data interface portion of OpenOffice have been done with inconsistencies WRT the rest of the packges. I think that the problem I reported in 4811 is specifically a data interface problem. I am also of the opinion that this problem might trigger the bug reported in 3045, but I am not very sure of this. >The reason why I originally assigned 4811 to Writer (besides that I >suppose part of the problem is there) is that investigations have to >_start_ there. Normally, we would need something like >"Database/Writer interaction" :). But we haven't, so we have to >decide for something ... Again, two separate issues. >Okay. I admit my argueing for the "Word processor" component may have >been inspired by the technical background. From a user's perspective, >it's likely that user's recognize this as "Database Access" bug >instead. So I would agree to change the component field of 3045 back >to "Database Access". >But _then_ 4811 really is a duplicate, isn't it? :) I'm afraid not ... Sincerely yours, Emmanuel Charpentier
I haven't any possibility to delve into source code ; therefore, since IssueZilla wants me to reassign this issue, I am assigning it to the owner of DB component code.
See recent comment in Issue # 4767 : I feel more and more that both issues are related to a "middleware" problem in the new-and-shiny data interface.
Emmanuel, I'm still not convinced that this issue is not the same as issue 3045, but I didn't yet find the time to react on your agueing. For the moment, I grab this issue for myself (it was assigned to you to ask to to give some more info, not to dig into the source :). Frank
Fixed in next release.
closing
change subcomponent to 'none'