Hi, Quite a number of these are used in various unit-tests. We added only a few for test-integration alone to verify certain types of regressions, all others either were there before or added for unit-tests. So the gain in checkout-speed is likely not big unless you check for each file if it is used in some unit-test or not before excluding it.
Also there are unit-tests which iterate over all these files to detect any upcoming problem early, e.g. the TestBiffViewer and related ones, while these would still run, they would obviously test less then. Dominik. On Sun, Nov 13, 2016 at 10:09 PM, Javen O'Neal <[email protected]> wrote: > Can all of the test-data files be excluded from the releases and svn > checkout? This might be less effort to maintain than defining two sets: > integration-tests and files used for unit tests which are required to > generate poi-ooxml-schemas. > > The test-data resources take the longest portion of an svn checkout, and > could be lazy downloaded when running a test. For people wanting to merely > browse the source code, these files are not necessary. > > On Nov 13, 2016 11:12 AM, "Andreas Beeker" <[email protected]> wrote: > > > The idea was to include only part of the test-data which is necessary for > > the test > > excluding the integration test AND have a special corpus for > > integration-tests, > > which can be downloaded on demand. > > > > The motivation was to keep the releases smaller. > > > > For the second part, it would be nice, if we have different collections, > > e.g. poi-basic (the additional files, which are currently used for > > integration test), > > tika (tika office corpus???), gov-docs, common-crawl, common-crawl-excel, > > common-crawl-10gb (only 10gb) > > > > Andi > > > > > > > > On 13.11.2016 18:05, Dominik Stadler wrote: > > > Hm, we are including the test-data directory in the sources as far as I > > > see, so you should be able to run test-integration when you download > just > > > the source-package, or do I miss something here? > > > > > > Dominik > > > > > > On Sun, Nov 13, 2016 at 4:57 PM, Javen O'Neal <[email protected]> > wrote: > > > > > >> +1 for this idea. > > >> > > >> Possible solutions: > > >> 1) Publish the commands for a sparse svn checkout on the website. It > > looks > > >> like Subversion doesn't have a simple "svn checkout > > >> https://svn.apache.org/repos/asf/poi/trunk poi --exclude-dir > > test-data", > > >> but we could get the same behavior with a checkout immediates, > checkout > > >> infinity awk listdir exclude test-data. > > >> This could he packaged into a bat/shell script, ant target, or Gradle > > >> target. > > >> 2) retree test-data to be a sibling of trunk. We would need have some > > way > > >> of pinning test-data so that old releases could be run against these > > >> documents without breaking. > > >> 3) Migrate away from asking users to check out the source using a > > >> Subversion client, using Gradle to perform this checkout instead > > (solution > > >> 1). > > >> > > >> On Nov 13, 2016 5:41 AM, "Andreas Beeker" <[email protected]> > wrote: > > >> > > >>> Hi, > > >>> > > >>> our test corpus is constantly growing and I think this is good, as > this > > >>> covers the edge-cases in the integration tests. > > >>> > > >>> But I wonder if we need to include those files in the releases, e.g. > we > > >>> could make those files downloadable in case > > >>> a users executes test-integration. Or maybe we find a way to have a > > >> common > > >>> corpus with tika ... but it should be > > >>> easy to download/test those with/-in the ant/gradle scripts. > > >>> > > >>> What do you think? > > >>> > > >>> Andi > > >>> > > >>> > > >>> ------------------------------------------------------------ > --------- > > >>> To unsubscribe, e-mail: [email protected] > > >>> For additional commands, e-mail: [email protected] > > >>> > > >>> > > > > > > > > --------------------------------------------------------------------- > > To unsubscribe, e-mail: [email protected] > > For additional commands, e-mail: [email protected] > > > > >
