|
From: Colin B. <col...@bi...> - 2002-09-29 05:31:41
|
Hi Scott, > First, I think current CVS is ready, as it's a definite improvement ove= r > the last released version for me. When I tried 0.9.6, it took a very > long time to connect (I think it was trying to load all of the metadata > right away, correct? I'm connecting over the Internet to a large Oracle > database, so that's not a good idea.) It's better now. Also, I think th= e > old one leaked cursors - I got maximum cursors exceeded messages when > browsing metadata. Theres a release candidate coming out in the next couple of days (when I = get=20 the time). Apart from finishing the documentation and bug fixing this sho= uld=20 be the same as the final release. > That doesn't mean I think it's perfect, though. I've got some > problems/interface comments/feature requests: Unfortunately its never going to be perfect, but the more feedback we get= the=20 closer we can make it :-) > - Typing "./squirrel-sql.sh" doesn't work for me right in the dist > directory unless I set INSTALL_PATH=3D`pwd`. Was this supposed to be > substituted in the ant file? If so, it didn't work. The Izpack installer sets it up during the installation. Its done this wa= y so=20 that the script can be run when the working directory isn't the Squirrel=20 directory. > - Ideally, just a "java -jar squirrel-sql.jar" would work, so I wouldn'= t > need the shell script/batch file at all. I think you can have classpath= s > in the meta-inf. If not, the other jars would have to be included in > that one. This would be better because batch files/java pop up > unnecessary terminal windows on Windows. Clicking on the jar file > directly launches javaw, which doesn't have that problem. The squirrel-sql.jar file has been executable for the last few betas. Tak= e a=20 look in the supplied batch file for other ways of executing Squirrel. The= =20 only issue with using the executable jar file is you don't get the conten= ts=20 of the %CLASSPATH% variable appended to the Squirrel class path and so yo= u=20 don't have access to any JDBC drivers in the %CLASSPATH%. > > - It's slow when browsing metadata. It would be nice if it cached more > (column-level information particularly) so looking at it a second time > would be quick. Info for the object tree is cached while the object is selected. I.E. if = a=20 table is selected and you display the column info, then the primary keys = and=20 then return to the column info the second display of the column info is f= rom=20 a cache and is much faster. It was a deliberate design decision on my par= t so=20 that changes to the object would be seen the next time you displayed it. = I=20 assumed that retrieving the metadata for a single tab would be a relative= ly=20 quick operation. It never occured to me that people would be connecting t= o=20 databases across slow connections. I'll make a change in 1.2 to cache the= =20 metadata when its first displayed and leave it cached until the object tr= ee=20 is refreshed. > - IMHO auto commit should be off by default. Or at least there should b= e > a preference for it per-database alias, not per-session, so I don't nee= d > to toggle it every time I connect. In the SQL tab on the New Session properties dialog (File/New Session=20 Properties) you can specify whether new sessions default to autocommit on= or=20 off. > > - I'm glad to see it doesn't block waiting for a database response. Thi= s > was what started me looking for an alternative to the Oracle 8i SQL > Worksheet to begin with. Killing the program when I forgot a join > condition was not pleasant. IMHO one of the best enhancements to 1.1 over 1.0 was Johan Compagners=20 threading code. Its made Squirrel much more useable. > - The metadata views show a lot of useless columns. Ones that are > consistently <null>: schema on PostgreSQL (even 7.3b1 - I guess they > haven't updated their JDBC drivers), table_cat on Oracle. And some > things not really intended for humans (data_type; the numbers mean > nothing to me; I just want type_name). If these all went away, it would > be easier to find what I'm looking for. Originally I only showed the _important_ columns in metadata but a couple= of=20 people were getting SQL errors when retrieving only some of the columns. = The=20 quick fix was to retrieve all columns but I wasn't able to track the prob= lem=20 down. It was one of those intermittent issues. Longer term I intend to ha= ve a=20 GUI whereby you can select the columns (and the sequence) you want for ea= ch=20 metadata tab. Shorter term I could put up a session property called somet= hing=20 like "Only show important metadata" true/false. - If you could come up wi= th a=20 better name for it I might put it in 1.1 > * I tend to work with a couple queries at once and refine them. It'd > be nice if a new tab replaced the current one unless I hit ctrl-T (as i= n > Mozilla). This way, older versions of the current query would go away. > > * And also, if the SQL was per-tab as well, so I could refine a > couple queries in parallel without doing as much copying and pasting. Better handling of the tabs is on my todo list but as yet I haven't reall= y=20 worked out what I want it to do. > * I'd like to see each SQL tab go into the top-level tab bar instead > of the one "SQL" one and subtabs of that. Maybe not conceptually quite > as clean - "Objects" would be different from all the others then - but > less clutter. You'd have a hard time convincing me on this one. I think it would make i= t a=20 lot messier. > > - I hate the MDI interface/subwindows. I usually only work with one > database at once - I only want that window up. Even when I do work with > two databases, I'd prefer they be entirely separate windows. The connec= t > stuff could be a dialog box that pops up when you first start the > program and when you hit ctrl-N or something (new connection). As I said recently to somebody, this is a religious issue. I personally l= ike=20 the MDI interface and hate the SDI one, this was the main reason I stoppe= d=20 using Netbeans. A possibility for the future when the code has been clean= ed=20 up would be making it an option as it is in the current version of Netbea= ns.=20 You specify whether you want an SDI or an MDI interface.=20 > - It would be nice if the statement division stuff was smarter. I'd lik= e > to be able to type in PL/SQL blocks without changing the statement > seperator. If it could do everything as the Oracle SQL Worksheet 8i doe= s > (not the 9i; blank lines in "create table" commands confuse it; grr), > I'd be really, really happy. Changing the statement separator and beginning of line comment gets round= =20 these issues so I'd look at this as being low priority. > - Feature request: Oracle query execution plans, like their "SQL > Scratchpad" in the "Enterprise Manager Console" has. SQL Scratchpad > really annoys me, but as is I've gotta use it occasionally. "explain > plan for" dumps all this info into a table, but the SQL Scratchpad is > the only thing I've seen that does a decent job of formatting it. (The > only second is this utlxplp.sql thing that does text-based formatting.) Sounds like something that could be put into the Oracle plugin. Its also = not=20 something I'd probably do myself (I'm no Oracle expert) so if you are=20 interested in enhancing Squirrel this could be a good starting point.=20 Squirrel is starting to be used in a lot of Oracle shops and theres a lot= of=20 oracle specific functionality that could be added to the plugin. Let me k= now=20 if you are interested. > If I'm feeling ambitious later and have time, I might try to implement > some of my suggestions. No promises. Squirrel is turning out to be bigger than Ben Hur so any help you can giv= e=20 would be appreciated. As I said the oracle plugin needs a lot of work and= I=20 would be extremely happy if you were to have a look at that. --=20 Colin Bell http://squirrel-sql.sf.net |
