shawn

Do you know of a way to consume a generic version number from these URLs?

when attempting to post a reply a moment ago I received a MySQL error. An error was encountered -------------------------------------------------------------------------------- Unable to connect to MySQL and select database. MySQL reported: Can't connect to local MySQL server through socket '/var/lib/mysql/mysql.sock' (2) The error occurred on line 44 in [path removed]/mysqli.php This is common when mysql exceeds the number of allowed connections, the hdd is failing, during updates, or when the amount of memory (or other resources) allocated to MySQL is insufficient. I don't know what level of control you have over this, but if you are interested in free hosting (on resource-safe servers), I will provide.

That would satisfy most of my needs.

I agree on both of these. Note that the "variables" window will need the title bar adjusted to include the appname. I've tried running two instances in order to compare jobs and what I learned was - folks, don't do that. Changes made to the second instance before closing it were lost to the first instance. Sigh.

I needed the ability to randomly select a source from a site the other day, in order to not fixate the download to one specific source and overburden it (not that that's very likely), but it would have been easier if there were a "random" function within Ketarin. I imagine it would operate in several different syntaxes, based on the source and parameters fed to it. Here's simplest form: {varname:random} This would parse and return one of the lines from the {varname} variable. Parsing would be performed by default at carriage returns. Another: {varname:random:delimiter} This one parses {varname} in the same fashion, but at the delimiter specified, instead of at carriage returns. Another: {random:list:var1:var2:var3:var4:var5:var6} This syntax would randomly return one of the variables named within it (note the lack of a 'varname' prefix). Another: {random:range:3000:3006} Would randomly choose a value, inclusively, between 3000 and 3006. Another: {random:varname} This is the one that would be most applicable to me immediately. This one changes the behavior of a RegEx or Start/End variable to instead parse for multiple results of the RegEx or Start/End combo and return a RANDOM result from the matches. Yes, I understand this would be freaky and probably require a large rewrite of the variable handling stuff, but it would let us use a truly random source when available - instead of the first result. Imagine using it with this regex, for example: <option value="([^"]+)">

I use the following: {swebsite} - website URL (a replacement for the "website" box on the information tab so that I can access it programmatically) {schangelog} - changelog URL for this program, often needs to be computed based on other input, such as the version or date {snotes} - long text description of the program or other oddities of it, such as prerequisites or installation instructions {version} - duh {sdownload} - if present, the partial or complete URL to download {filepath} - a fragment used in building {sdownload} {filestart} - a fragment used in building {sdownload} {filemiddle} - a fragment used in building {sdownload} {fileend} - a fragment used in building {sdownload} {fileext} - a fragment used in building {sdownload} {filepattern} - a regex pattern used in building {sdownload} {tag} - (or {sftag}, {mstag} or similar) is a specific key to obtaining the correct URL from a search or index page on the source site. {aversion} - fragment or 'step' in building the correct {version} string {bversion} - fragment or 'step' in building the correct {version} string {cversion} - fragment or 'step' in building the correct {version} string, repeat ad nauseum until the version is formatted correctly, but rarely is it required to iterate this more than 3 or 4 times.

I can cope with that - but I hope the others will forgive me.

I have a request. I know it's kinda late (already being beta), but maybe not too late in the game. The entity-encoding of the template is (IMHO) the wrong way to do this. It would be far more portable and externally (such as thru batch, shell, vbs and so on) if you used XML CDATA blocks to store the exact template, so no interpretation, reformatting or entity-encoding is required externally when automating data manipulation.

Here, too. Whew. I really didn't want to have to go try to figure out what the heck was wrong.

All my SF apps are broken, too. Anything using "http://sourceforge.net/" as the referer is broken. When I have some time later I'm going to modify the referer to use the specific project files page, and we'll see if that makes a difference.

bug report (1.5b3): Somewhere in the last few iterations the custom columns can no longer include the following as the variable source: {property:HttpReferer} The referer shows in the properties window, and in the exported XML, but not in the custom column as assigned. Also, all my SF stuff just failed. Maybe the Referer value isn't being processed correctly right now at all?

No? Why not create a new textual variable that consists of {f:...} variables? The file date will actually correspond to the "URL date". The problem is, that method works successfully only if the filename doesn't depend on the variable value. It can't parse the {f:...} stuff to compose the values to create the filename. Creating a test case...hm, now it's working. Aaarrrrgggghhhh. Hmm. Maybe it only raises that error if the file hasn't existed yet?

Check out "start.exe". That provides options for shelling out of process. You would have to consume then follow the link as a two-step process. If possible, consider using an RSS feed or tag page, when available.

Yep. But also remember that the database itself also contains a GUID (a database is considered as an "author"). This is the authentication for updating the existing application. If you throw your database away, you lose the GUID and cannot update any applications that have been uploaded from that database. I understand that...but my point is, the GUID is completely different. It's not the same "app" anymore. No more than if I were to completely replace the engine in my car -- it might have many of the same structural details, but it's not the same thing. The deciding factor - the "uniqueness" -- is determined by the GUID, right? if that really is the only determining factor, then there's no reason that the app I tried to share wouldn't be allowed: It had a different GUID.

Bug report: Somewhere during the recent alphas and betas, Batch scripts that included ">" within them had the greater than character and everything to the right of it removed. This was around 9/26, if that helps at all. I didn't make incremental backups of the db, so can't tell exactly which version did it.

Oh, on that note - Flo - if you decide to add cookie handling abilities, please provide a new time/date function to return the unix timestamp. And speaking of date functionality.... Feature request: Would it be possible to add (somehow) the ability to access the file or URL timestamp for storing within a variable ({version}). In my specific scenario, I'm trying to assign a version to a file that doesn't really have a published version number anywhere (CrucialScanner.exe). The only unique data is that of the Last-Modified header returned in the request...and while I know that the {f:...} methods can be used for file naming, I don't see any equivalent functionality for variable assignments.

That's a perfect example of why you need the "run command after download" functionality. Parse the URL from the download source (mediafire or whatever) into a variable. Download it, or not (and download something uber-tiny like an icon file) then use the following command in the post-download command: java.exe -jar jdownloader.jar -a "{yourUrlVaraibleName}" Nifty, eh? It's a 406 error (not acceptable), so it's probably a referer/cookie/user-agent issue. Set a known good user-agent and referer in the advanced tab, and try again. Doing so doesn't work for me, so it's most likely a required cookie is missing. Checking...yep. Visiting the site assigns a cookie with a timestamp: lastVisit=1285751139 In order to download it, we'd need to be able to assign a cookie, or at least ensure js-based cookies are scripted correctly and values assigned during subsequent page loads. Setting a cookie to the current unix timestamp would probably work (it does in my Perl tests).

True, but if you create a new app profile using the "old" template code, then export it, you'll be able to see the source is extremely simple (simply encoded text of the template) so migrating that section to all the rest of the similar templates shouldn't be too difficult. Back it up a few times first before you do, though, since an error could be Egon Spengler "bad". I got the same thing during most of my tests - use lowercase and the attribute name "Guid" not "guid" or "GUID". That seems to work for me.

If you do that, uploading a new version to the database is not possible. You need to keep the original GUID if you want to refer to the same application. Understood, but if the GUID is the determining factor (and it's a DIFFERENT application - the original was Quicktime Alternative, this one is Quicktime), it shouldn't matter, right? I don't know how this variable is defined. There should not be any change in the behaviour of variables though. I can't replicate the issue right now, but {filepath} was defined as a "textual content" variable, and was empty (the default). When the URL was created for the version check, {filepath} was not replaced with "". Can you please change this so that we *can* add additional placeholders, then prompt us what to do with them - and then a prompt to "would you like to use the following default value for all missing {whatever} placeholders, or be prompted for each individual app?" This would greatly ease my mind in using the prototemplate functions. Yep. Works for me.

Oh, oh...and the ability to right-click over a specific "cell" in the grid (like 'target' or 'custom column 2') and copy the contents of JUST that cell. This would be a HUGE time-saver for me when something breaks.

Hm. The behavior change for templates has broken all my previous templates. Now it requires a GUID for all templates, even if the template is not intended to be a prototype (though I can respect how that really should be the default behavior). (also note: GUID must be in lowercase!) I found a more reliable method of getting the current QuickTime Player (filehippo has been broken on this for several weeks), and wanted to share it --- but it's not working the way I expect it to. The online database (even after changing the GUID a half dozen times, deleting and reimportant and stuff) still tells me I'm attempting to upload "QuickTime Alternative" and asks me to rename it. Also, when I do create a new app based on one of my modified templates (Guid added to sfnet template), the resulting search doesn't work, complaining of a 404 when accessing "hxxp://sourceforge.net/projects/{sftag}/files/{filepath}" (which should resolve to "hxxp://sourceforge.net/projects/dqsd/files/"). The URL works, and the SF referer is in place. Any thoughts? My guess is that there is a behavior change in the new version that no longer correctly returns an empty string within "{filepath}" since removing that portion of the URL box fixes it. Or maybe the variable replacement function for this field no longer iterates until fully resolved?

Hey, here's a feature request along these lines. How about a styling option to make empty or invalid values in custom columns more visible. Like, a red background or something. Obviously, this is one of those things you'd want to assign on a column by column basis (probably a checkbox next to the column/variable assignment field). OR a special regex pattern to attempt to match. For example, if CustomColumn1 is set to the following Error Format "[\d\.\-]+" then a value of "3.5 Beta 2" would be highlighted red. Also, I'm really feeling my hands tied with only two custom columns. Can we have some more, please? I could easily see myself using 4, and might need more in some situations. Pretty please?

Status bar change is perfect. The dropdown for "use the following variable as indicator for changes" should always include "version", even if it's not a defined variable, since this is the behavior of certain native functionality (such as filehippo and pad parsing). How does the prototype template functionality work? I understand that it needs a GUID - easy enough (in DQSD it's "guidgen"). I guess what I need to know is, is the logic to handle proto-templates essentially "if the xml file has a 'placeholder' AND a GUID, it's a prototemplate - but if it only has only one or the other, it's a normal xml app profile." This is important because I currently use exported app profiles to copy and **update** the settings from one system to another. If the GUID is non-unique, or generated randomly during import instead of preserved for some reason, then it'll definitely raise issues.

CybTekSol, there are times when the version number returned by FileHippo is not formatted the way I want it to be, so I have to use a two-off variable to correctly obtain the version info, then format it. It's a pain, but necessary. If only one step is required, I can do it in the naming box, but sometimes more complete formatting is required. It occurs to me that I might be able to use a new version-type variable ({myversion}), with a value like: {version:multireplace:...} And then use {myversion} within the file naming box. That ought to simplify things a bit.

I have {version} as a custom column indicator, so even before the errors are raised for a job, I can tell immediately that those returning "<!DOCTYPE..." are broken, one way or another. Under the "blank" option, these fields would return empty, so the {version} column would be blank. Simple enough to find and resolve. Assuming resolution is possible.