webmockrnow supports thehttr2library, in addition tohttrandcrul. Note that you'll see different behavior fromhttr2relative to the other 2 http clients because it turns http errors (http statuses 400 and above) into R errors (#122)webmockrcan now mock async http requests withcrul(w/crulv1.5 or greater). no change was required inwebmockrfor this to happen. a PR was merged incrulto hook intowebmockr. there's no support for async inhttras that package does not do any async and no support inhttr2becausereq_perform_paralleldoes not have a mocking hook as doesreq_perform(#124)
to_return()supports returning multiple responses to match many requests to the same matching stub. however, the internals were broken for this, but is now fixed (#115) thanks @kenahoo for the report- matching stubs with specifying a request body to match on (e.g.,
stub_request('post', 'https://httpbin.org/post') %>% wi_th(body = list(a=5))) was not working in some cases; internal matching logic was borked. now fixed. (#118) thanks @konradoberwimmer for the report - The
statusparameter into_return()was documented to accept an integer, but it errored when an integer was passed (e.g.,to_return(status=200L)). This bug is now fixed (#117) thanks @maelle for the report
- Config options changes (see
webmockr_configure()). Three options that were presentg but not implemented are now removed:show_body_diff,query_values_notation,net_http_connect_on_start. One option that was present but not implemented yet is now implemented:show_stubbing_instructions(#27) (#120)
StubCounteradded to pkgdown docs page at https://docs.ropensci.org/webmockr/reference/StubCounter.html (#119) @maelle
- change to
UriPatternto make sure regex matching is working as intended (#114) thanks @kenahoo
enable()and theenable()method on theAdapterR6 class gain new parameterquietto toggle whether messages are printed or not (#112)
- to re-create http response objects for both httr and crul we were using the url from the request object; now we use the url from the response object, BUT if there is no url in the response object we fall back to using the url from the request object (#110) (#113)
- improve docs: add further explanation to manual files for both
to_raise()andto_return()to explain the differenc between them and when you may want to use them (#100)
- to support vcr being able to recreate httr objects fully (see github issue ropensci/vcr#132) we needed to handle additional parts of httr request objects: fields and output - with this change vcr should return objects much closer to what real httr requests return (#109)
- bug fix + improvement: fixes for simple authentication -
wi_th()now supportsbasic_authto mock basic authentication either withcrul::auth()orhttr::authenticate()(#108)
- Gains ability to define more than 1 returned HTTP response, and the order in which the HTTP responses are returned. The idea is from the Ruby webmock library, but the implementation is different because the Ruby and R languages are very different. You can give more than one
to_return()one creating a stub, or if you want to return the same response each time, you can use the newtimesparameter withinto_return(). As a related use case (#31) you can mock http retry's using this new feature (#10) (#32) (#101) - Gains new function
webmockr_reset()to be able to reset stub registry and request registry in one function call (#97) (#101) - Gains support for mocking simple authentication.
wi_th()now acceptsbasic_authin addition to query, body, and headers. Note that authentication type is ignored (#103)
- change to how URI's are matched in
stub_request(): we weren't allowing matching URI's without schemes; you can now do that. In addition, webmockr can match URI's without the "http" scheme, but does not match if the scheme is "https". SeeUriPatternfor more (#102) - another change to how URI's are matched: now query params compared separately to the URI; note that regex not allowed in query params (#104) - And now query parameters are compared with the same code both when regex uri is used and when it is not (#107)
- URI matching for stubs is now done only on the URI's themselves; that is, query parameters are removed before comparison, so only the base url with http scheme, plus paths, are compared (#107)
- wasn't sure
write_disk_pathbehavior was correct when using httr, seems to be working, added tests for it (#79) - values for query parameters given to
wi_th()are now all coerced to character class to make sure that all comparisons of stubs and requests are done with the same class (character) (#107)
- fix for
uri_regexusage instub_request(): no longer curl escape theuri_regexgiven, only escape a non-regex uri (#106)
- change to
CrulAdapter: do not usenormalizePathon thewrite_disk_pathpath so that relative paths are not changed to full paths - added tests for this (#95) (#96)
- new
Adapterclass to consolidate common code for theHttrAdapterandCrulAdapterclasses, which inherit fromAdapter; not a user facing change (#87) - pkgdown documentation site gains grouping of functions to help the user navigate the package: see https://docs.ropensci.org/webmockr/reference/ (#93)
- now correctly fails with informative message when
write_disk_pathisNULLwhen the user is trying to write to disk while using webmockr (#78) - improve README construction; use html child for the details section (#81)
- fix matching stub matching for bodies when bodies are JSON encoded (#82)
- when vcr was loaded real HTTP requests were being performed twice when they should have only been performed once (#91) (#92)
- fix for
set_body()method in theResponseclass - handle cases where user writing to disk and not, and handle raw bytes correctly (#80) - fix to
to_s()method inStubbedRequestclass - was formatting query parameters incorrectly (#83) - fix to
BodyPatternclass to handle upload objects in a list; related issue fixed wherewi_th()parameterbodywas not handling upload objects (#84) (#85) - httr requests were failing when vcr loaded, but with no cassette inserted; fixed
handle_request()to skip vcr-related code unless a cassette is inserted (#86) (#88)
webmockrnow supports mocking writing to disk. TLDR: see?mocking-disk-writingto get started - That is, both of the major high level http clients in R, crul and httr, support writing directly to disk (rather than the user manually getting the http response and writing it to disk). supporting this required quite a bit of work, both in code and in thinking about how to support the various scenarios in which users can find themselves when dealing with writing to disk - Please get in touch if you have problems with this (#57) (#76)- gains
request_registry_clear()method to easily clear all requests in the request registry (#75)
- better docs for R6 classes with R6 support in new roxygen2 version on cran (#77)
- httr simple auth was being ignored - its now supported (simple auth with crul already worked) (#74)
- fix to handle raw responses that can not be converted to character, such as images; needed due to issue ropensci/vcr#112 (#72) (#73)
- fix link to http testing book, change ropensci to ropenscilabs (#67)
- fixes to request matching: single match types working now (e.g., just match on query, or just on headers); in addition, header matching now works; added examples of single match types (#68) (#69)
- fix stub specification within crul and httr adapters; typo in setting headers (#70)
- underscore methods
to_return_()andwi_th_()are defunct (#60) (#64)
to_return()gains parameter.list(#60) (#64)
- typo fixes (#62) thanks @Bisaloo !
- improved the print method for stubs, found in
StubbedRequest, to have better behavior for very long strings such as in headers and bodies (#63)
- fix date in mocked
httrresponse object to match the date format thathttruses in real HTTP requests (#58) (#61) via ropensci/vcr#91 - fix response headers in mocked
httrresponse objects.httrmakes the list of headers insensitive to case, so we now use that function from the package (#59) (#61) to_return()andwi_th()drop use of thelazyevalpackage and fall back to using the simplelist(...)- fixes problem where creating stubs was failing withintest_that()blocks due to some weird lazy eval conflicts (i think) (#60) (#64) thanks @karawoo !
- returned mocked response headers were retaining case that the user gave - whereas they should be all lowercased to match the output in
crulandhttr. now fixed. (#49) thanks @hlapp - returned mocked response headers were not all of character class, but depended on what class was given by the user on creating the stub. this is now fixed, returning all character class values for response headers (#48) thanks @hlapp
- skip tests that require
vcrifvcris not available (#53) - internal change to crul adapter to produce the same http response as a new version of crul returns - adds a
response_headers_allslot (#51) (#54)
- make
request_registry()andstub_registry()print methods more similar to avoid confusion for users (#35) - update docs for
enable/disableto indicate thatcrulandhttrsupported (#46) (related to #45) - wrap httr adapter examples in
requireNamespaceso only run when httr available - clean up
.onLoadcall, removing commented out code, and add note about creating adapter objects does not load crul and httr packages
- fix to
enable()anddisable()methods. even thoughhttris in Suggests, we were loading all adapters (crul, httr) withstopwhen the package was not found. We now give a message and skip when a package not installed. In addition, weenable()anddisable()gain anadapterparameter to indicate which package you want to enable or disable. Ifadapternot given we attempt all adapters. Note that this bug shouldn't have affectedvcrusers ashttris in Imports in that package, so you'd have to havehttrinstalled (#45) thanks to @maelle for uncovering the problem
- Added support for integration with package
httr; seeHttrAdapterfor the details;webmockrnow integrates with two HTTP R packages:crulandhttr(#43) (#44) - Along with
httrintegration is a new methodhttr_mock()to turn on mocking forhttr; and two methodsbuild_httr_responseandbuild_httr_requestmeant for internal use
- Added support for integration with package
vcr(now on CRAN) for doing HTTP request caching
- New function
enabled()to ask ifwebmockris enabled, gives a boolean wi_th()gains new parameter.listas an escape hatch to avoid NSE. examples added in thewi_thman file to clarify its use
- matching by request body was not supported, it now is; added examples
of matching on request body, see
?stub_request(#36) - make sure that the adapter for
crulhandles all types of matches (#29) - removed all internal usage of pipes in the package. still exporting pipe for users (#30)
- fixed internals to give vcr error when vcr loaded - for future release with vcr support (#34)
- require newest
crulversion
- Error messages with the suggest stub were not giving bodies. They now give bodies if needed along with method, uri, headers, query (#37)
- Fixed
Responseclass that was not dealing with capitalization correctly
- New function
to_raise()to say that a matched response should return a certain exception, currentlyto_raiseaccepts error classes from thefauxpaspackage (#9) - New function
to_timeout()to say that a matched response should return a timeout. This is a special case ofto_raiseto easily do a timeout expectation (#11) - New function
request_registry()to list requests in the request registry (#23) - package
crulmoved to Imports from Suggests as it's the only http client supported for now. will move back to Suggests once we support at least one other http client webmockr_configure()changes:turn_onhas been removed;allow_net_connectandallow_localhostwere ignored before, but are now used and are now set toFALSEby default; fixed usage ofallowwhich now accepts character vector of URLs instead of a boolean; the following correctly marked as being ignored for now until fixednet_http_connect_on_start,show_stubbing_instructions,query_values_notation,show_body_diff(#19) (#21)webmockr_disable_net_connect()now accepts anallowparameter to disable all other connections except those URLs given inallowwebmockr_net_connect_allowed()now accepts auriparameter to test if a URI/URL is allowed
- Fixed printed stub statement when printed to the console - we weren't including headers accurately (#18)
- Added examples to the
stub_registry()andstub_registry_clea()manual files (#24) - internal methods
build_crul_requestandbuild_crul_responsemoved outside of theCrulAdapterclass so that they can be accesed likewebmockr::in other packages enable()anddisable()now return booleans invisibly- General improvements to documentation throughout
- Added linting of user inputs to the
to_return()method, and docs details on what to input to the method - Added linting of user inputs to the
wi_th()method, and docs details on what to input to the method
- Fixed option
allow_localhost, which wasn't actually workin before (#25)
webmockr_enable()andwebmockr_disableare now defunct. Usewebmockr::enable()andwebmockr::disable()instead
- Released to CRAN.