3 ms·

C database connection pool library

57 points by sovande 15y ago

signa11 15y ago
i have a fundamental question regarding this: instead of a pool of connections, is it possible to have an event-loop based aysnc-io in place ? this is with little experience with db's...
abcd_f 15y ago
Good presentation (among other things).
atjoslin 15y ago
Great presentation, and implementation looks good too.
yarpa 15y ago
Showing example code on the home page is a great idea, having the small thumbnail with the expand icon made it easy to jump in and see what their API looks like. Going to give this library a try soon :)
saurik 15y ago
FWIW, I went to the website, clicked Documentation, and when nothing changed on the site except the button I started to wonder if the site was offline or my web browser had locked up. Then I clicked another link, noticed my scrollbar changed size, and realized you were updating content below the bottom of my screen due to the unusually enormous header.
- tommi 15y ago
  I like the design with the big header, but the page could scroll page when a section link is clicked.
zxoq 15y ago
Look good, but does it work on Windows? It's not on the operating system list. If it does, I might very well use this in the future. Edit: Saw it was GPL. Well it looks good at least.
jrmg 15y ago
Can I use libzdb in my iOS or Mac OS X app? Yes[...] Except it's GPLed, so it can't be used in any closed-source apps, or any apps that you want to distribute on the App Store, without an explicit exception. If you're not married to the license for philosophical reasons (which I would respect, it's your right), I'd suggest a change to something simple like BSD. This would, of course, allow someone to take your code, change it, and not release their changes, but it would also make it usable to the majority of programmers.
- zxoq 15y ago
  Indeed, I was excited until I saw the GPL. It really makes the library quite useless, unfortunately. :(
- Argorak 15y ago
  It doesn't have to be BSD, but at least relaxing the license to LGPL would be fine.
  - jrmg 15y ago
    Actually, no, that's a common misconception. The LGPL requires that the end-user be able to replace the library with an API-compatible library and still run the program. That's arguably impossible with an iOS device, and certainly impossible if you don't provide at least object files that can be relinked. Even with object files (or source) they'd have to pay to register as an iOS developer. I suspect the majority of developers using the LGPL don't realise, or perhaps even agree, with this intent, so you're probably unlikely to be sued, but it's what the license demands.
    - Argorak 15y ago
      I didn't say that using the LGPL would give you the same possibilities as BSD. Sorry for the answer that only half fitted the question, in context of iOS and App Store, you are right. If the author wants to use a GPL license, its his decision. But using a viral license in a support library is harmful in any case, as it destroys all ability to use it in a program that incorporates other pieces of code that might not be compatible with the GPL.
andrewcooke 15y ago
have you considered a varargs call for setting parameters with type codes? so something like PreparedStatement_setArgs(p, 's', "a string", 'i', 42, 'f', 3.1415926); would set arguments 1-3. i know it's not "OO", but i have done something similar in an internal lib for work and it's very useful and easy to use. [edit] the number of parameters is assumed equal to the number of '?' in the query. i also have something similar for reading results from the ResultSet, but that required pointers and a count, and was much more likely to lead to errors.
- hauk 15y ago
  > have you considered a varargs call for setting parameters with type codes? Varargs are supported when building a prepared statement, but not for setting wildcards values. I see what you mean and how the client code can be made more compact. Still, I believe setting one by one value by explicit type may be better for readability, is less error prone and as you say, is more OO and orthogonal.
  - andrewcooke 15y ago
    just trying to make it easier to use... one more thing that i assume must work this way, but you don't seem to have documented, is how to test for exceptions. i guess SQL_Exception (a return value, right?) is false when there's no error? related: do you have any examples that include all the error handling that would be necessary in a real app? how do you display the exception details (it seems to be an opaque type) to the user? perhaps i am missing something here (can't even find exception in the clickable API image)? (wish i had known about this a few months ago as it would have saved me some work - thanks).
    - hauk 15y ago
      Thanks. The documentation for exceptions may be some clicks away, but it's there. Here's a direct link. http://www.tildeslash.com/libzdb/api-docs/Exception_8h.html http://www.tildeslash.com/libzdb/api-docs/Exception_8h.html Exceptions are a "true" exception implementation based on the Except code from David Hanson's excellent CII book. This makes it possible to write more compact code without peppering your C code with return code tests.

57 points by sovande 15y ago

i have a fundamental question regarding this: instead of a pool of connections, is it possible to have an event-loop based aysnc-io in place ? this is with little experience with db's...

abcd_f 15y ago

Good presentation (among other things).

atjoslin 15y ago

Great presentation, and implementation looks good too.

yarpa 15y ago

Showing example code on the home page is a great idea, having the small thumbnail with the expand icon made it easy to jump in and see what their API looks like. Going to give this library a try soon :)

saurik 15y ago

FWIW, I went to the website, clicked Documentation, and when nothing changed on the site except the button I started to wonder if the site was offline or my web browser had locked up. Then I clicked another link, noticed my scrollbar changed size, and realized you were updating content below the bottom of my screen due to the unusually enormous header.

tommi 15y ago
I like the design with the big header, but the page could scroll page when a section link is clicked.

zxoq 15y ago

Look good, but does it work on Windows? It's not on the operating system list. If it does, I might very well use this in the future. Edit: Saw it was GPL. Well it looks good at least.

jrmg 15y ago

Can I use libzdb in my iOS or Mac OS X app? Yes[...] Except it's GPLed, so it can't be used in any closed-source apps, or any apps that you want to distribute on the App Store, without an explicit exception. If you're not married to the license for philosophical reasons (which I would respect, it's your right), I'd suggest a change to something simple like BSD. This would, of course, allow someone to take your code, change it, and not release their changes, but it would also make it usable to the majority of programmers.

zxoq 15y ago
Indeed, I was excited until I saw the GPL. It really makes the library quite useless, unfortunately. :(
Argorak 15y ago
It doesn't have to be BSD, but at least relaxing the license to LGPL would be fine.
- jrmg 15y ago
  Actually, no, that's a common misconception. The LGPL requires that the end-user be able to replace the library with an API-compatible library and still run the program. That's arguably impossible with an iOS device, and certainly impossible if you don't provide at least object files that can be relinked. Even with object files (or source) they'd have to pay to register as an iOS developer. I suspect the majority of developers using the LGPL don't realise, or perhaps even agree, with this intent, so you're probably unlikely to be sued, but it's what the license demands.
  - Argorak 15y ago
    I didn't say that using the LGPL would give you the same possibilities as BSD. Sorry for the answer that only half fitted the question, in context of iOS and App Store, you are right. If the author wants to use a GPL license, its his decision. But using a viral license in a support library is harmful in any case, as it destroys all ability to use it in a program that incorporates other pieces of code that might not be compatible with the GPL.

andrewcooke 15y ago

have you considered a varargs call for setting parameters with type codes? so something like PreparedStatement_setArgs(p, 's', "a string", 'i', 42, 'f', 3.1415926); would set arguments 1-3. i know it's not "OO", but i have done something similar in an internal lib for work and it's very useful and easy to use. [edit] the number of parameters is assumed equal to the number of '?' in the query. i also have something similar for reading results from the ResultSet, but that required pointers and a count, and was much more likely to lead to errors.

hauk 15y ago
> have you considered a varargs call for setting parameters with type codes? Varargs are supported when building a prepared statement, but not for setting wildcards values. I see what you mean and how the client code can be made more compact. Still, I believe setting one by one value by explicit type may be better for readability, is less error prone and as you say, is more OO and orthogonal.
- andrewcooke 15y ago
  just trying to make it easier to use... one more thing that i assume must work this way, but you don't seem to have documented, is how to test for exceptions. i guess SQL_Exception (a return value, right?) is false when there's no error? related: do you have any examples that include all the error handling that would be necessary in a real app? how do you display the exception details (it seems to be an opaque type) to the user? perhaps i am missing something here (can't even find exception in the clickable API image)? (wish i had known about this a few months ago as it would have saved me some work - thanks).
  - hauk 15y ago
    Thanks. The documentation for exceptions may be some clicks away, but it's there. Here's a direct link. http://www.tildeslash.com/libzdb/api-docs/Exception_8h.html http://www.tildeslash.com/libzdb/api-docs/Exception_8h.html Exceptions are a "true" exception implementation based on the Except code from David Hanson's excellent CII book. This makes it possible to write more compact code without peppering your C code with return code tests.