Minor rewording of 2.6 to make it easier to understand. #342

viktorklang · 2017-03-05T20:53:01Z

Addressing #321 with minimal changes. (No changes in spec semantics)

@reactive-streams/contributors Please review and vote.

ktoso · 2017-03-05T20:57:47Z

Wanting to look at the changeset shows me a 500 error page; does it work for you?

akarnokd · 2017-03-05T21:06:39Z

@ktoso Same here. I had trouble with #339 as well the past week, mostly since the Amazon outage. The patch links from the email work for me:

https://github.com/reactive-streams/reactive-streams-jvm/pull/342.patch
https://github.com/reactive-streams/reactive-streams-jvm/pull/342.diff

ktoso · 2017-03-05T21:07:20Z

Thanks, will have a look

viktorklang · 2017-03-05T21:09:26Z

Yeah, I have issues with GH as well, 500s all over the place.

ktoso · 2017-03-05T21:13:12Z

Uhm, if the change is only to add the , then I don't think it's helping enough.
The linked issue proposes a more wordy version, however we could find a balance between the current and the new one perhaps?

How about:

A Subscriber MUST call Subscription.cancel() when it is no longer interested in (or able to) receiving signals from its Publisher, even without the Publisher having signaled onError or onComplete.

viktorklang · 2017-03-05T21:16:14Z

@ktoso, it changes "the Publisher" to "its Publisher" and adds the comma.

ktoso · 2017-03-05T21:18:05Z

I see, the valid to its... still is not completely clear though, as witnessed by that issue. We don't use that wording anywhere else either AFAICS. A subscription can be valid, but this is the single case we used it for a Subscriber.

viktorklang · 2017-03-05T21:40:32Z

No, it is not about the Subscriber being valid, it is about the Subscription not being valid.

ktoso · 2017-03-05T21:41:58Z

Proves my point that it's easy to confuse the "it" then though?

viktorklang · 2017-03-06T08:54:45Z

@ktoso Very good point. See my second attempt.

ktoso · 2017-03-06T08:58:23Z

Hmm, very simple wording and I think it does indeed cover all the cases we mean here.
The extended intent section explains what we mean well now.

Sounds good to me 👍

viktorklang · 2017-03-06T09:52:07Z

README.md

@@ -139,8 +139,8 @@ public interface Subscriber<T> {
 | [:bulb:](#2.4 "2.4 explained") | *The intent of this rule is to make sure that Subscribers respect a Publisher’s [terminal state](#term_terminal-state) signals. A Subscription is simply not valid anymore after an onComplete or onError signal has been received.* |
 | <a name="2.5">5</a>       | A `Subscriber` MUST call `Subscription.cancel()` on the given `Subscription` after an `onSubscribe` signal if it already has an active `Subscription`. |
 | [:bulb:](#2.5 "2.5 explained") | *The intent of this rule is to prevent that two, or more, separate Publishers from thinking that they can interact with the same Subscriber. Enforcing this rule means that resource leaks are prevented since extra Subscriptions will be cancelled.* |
-| <a name="2.6">6</a>       | A `Subscriber` MUST call `Subscription.cancel()` if it is no longer valid to the `Publisher` without the `Publisher` having signaled `onError` or `onComplete`. |
-| [:bulb:](#2.6 "2.6 explained") | *The intent of this rule is to establish that Subscribers cannot just throw Subscriptions away when they are no longer needed, they have to call `cancel` so that resources held by that Subscription can be safely, and timely, reclaimed.* |
+| <a name="2.6">6</a>       | A `Subscriber` MUST call `Subscription.cancel()` if the `Subscription` is no longer needed. |


One question which comes to mind is if we need to be explicit here that this rule only applies as long as it doesn't violate any other rule (i.e. not calling Subscription methods after terminal state has been received etc.)

viktorklang · 2017-03-06T09:52:21Z

README.md

-| <a name="2.6">6</a>       | A `Subscriber` MUST call `Subscription.cancel()` if it is no longer valid to the `Publisher` without the `Publisher` having signaled `onError` or `onComplete`. |
-| [:bulb:](#2.6 "2.6 explained") | *The intent of this rule is to establish that Subscribers cannot just throw Subscriptions away when they are no longer needed, they have to call `cancel` so that resources held by that Subscription can be safely, and timely, reclaimed.* |
+| <a name="2.6">6</a>       | A `Subscriber` MUST call `Subscription.cancel()` if the `Subscription` is no longer needed. |
+| [:bulb:](#2.6 "2.6 explained") | *The intent of this rule is to establish that Subscribers cannot just throw Subscriptions away when they are no longer needed, they have to call `cancel` so that resources held by that Subscription can be safely, and timely, reclaimed. An example of this would be a Subscriber which is only interested in a specific element, which would then cancel its Subscription to signal its completion to the Publisher.* |


Is this example/clarification good enough?

viktorklang · 2017-03-13T09:16:28Z

@reactive-streams/contributors Yay or nay?

viktorklang · 2017-04-04T08:14:41Z

@reactive-streams/contributors Ping!

smaldini · 2017-04-11T14:32:52Z

@viktorklang @reactive-streams/contributors squash and merge ?

@viktorklang

* Repairs formatting issue of tables in spec README * Modifies rules 1.09 and 2.13 to mandate `java.lang.NullPointerException` be thrown. Updates the TCK, Spec and example implementations. * Fixes reactive-streams#211 by clarifying * Fixes reactive-streams#210 by removing 1.12 and repurposing its TCK checks for 1.09 * Clarifies the signalling sequence in the spec and adds TCK verification to ensure signal ordering is proper, also amends the examples to reflect the spec change. * Publish 1.0.0.RC2 fix reactive-streams#215 * Fixes reactive-streams#217 by including the examples project in the publish task * =tck minor test name fixup, it is a required test * fix reactive-streams#212 issue on spec 213 testing wrt Processor * RC3 release /w reactive-streams#222 fix * remove rule 1:12 (produce same elements to all Subscribers) This rule is in conflict with 1:11 which allows a Publisher to treat multiple Subscribers as either as unicast or multicast recipients. The verification of proper multicast behavior (which 1:12 specified) has been retained, the test methods renamed accordingly. * fix three left-over references to deleted rule 1:12 * Fixed wrong footnote reference in README.md * Addresses a couple of typos in the examples for AsyncSubscriber and SyncSubscriber * !TCK clarify what error publisher is + add better readme on what this method is + add better javadoc on this method - removes reference to old style spec annotation from readme + proposing to change method name to "createFailed..." as it is the wording used in the spec and reactive manifesto (footnote 1.1) + more info in tck/README that it is not legal to signal on* before sub Resolves reactive-streams#237, reactive-streams#235 * +tck reactive-streams#236 example subscriber whitebox tested, and whitebox fixed * add space to javadoc * +TCK verifyNoAsyncErrors now by default waits, fixes spec111 Resolves reactive-streams#239 * =tck general tck/readme.md cleanup so it matches current code / spec Resolves reactive-streams#99 Depends on reactive-streams#241 * Addresses PR review comments for reactive-streams#246 * Update CopyrightWaivers.txt * +tck explains createElement in more useful terms resolves reactive-streams#231 * +tck reactive-streams#232 explain which tests are mendatory to be "compliant" * Update SubscriberWhiteboxVerification.java Fixes Javadoc generation on Java8+ by having to manually qualify nested classes. * Fixes reactive-streams#233 by implementing support for triggered demand in in the SubscriberBlackboxVerification * Travis PR validation using both JDK 6 and 8 By validating on both JDKs we know the project even builds on 8, while not using features (classes) from JDK8 - so it's still usable for JDK6 projects. Resolves reactive-streams#254 * Small touchups to the TCK README.md * Release 1.0.0.RC4 * Cancel the subscription after receiving all of the pertinent emissions (reactive-streams#259). * Test that 'required_spec317_mustNotSignalOnErrorWhenPendingAboveLongMaxValue' completes in a timely manner for fully synchronous publishers (reactive-streams#259). * =tck untested spec308 rule method name adjusted * -tck rm undocumented and unused publisherReferenceGCTimeoutMillis method * update version to 1.0.0.RC5 * Updating documentation to reflect the current version: RC5 * update ref to 1.0.0.final * change 1.0.0.final to 1.0.0 and make sure OSGI manifest has the bundle version * OSGI fix * OSGI fix... * Disambiguate "processing elements" The document generally refers to "elements" as objects traversing a stream. I initially considered simply editing "processing elements" to read "processing components", but there's a section devoted to the definition of this, so better to link them. * Added per request of @viktorklang in reactive-streams#269 * add CC0 label to README * =tck reactive-streams#279 improve completion latch error message * Rename SyncSybscriber.foreach to whenNext * Update README.md Spelling of the company name is Red Hat, not RedHat. * I hereby represent [...] public domain [...] entirety of my contributions. Requested by @viktorklang. * Log test output events to the console * Remove "preview" qualifier from README. * Unbreaks TravisCI OpenJDK6 hostname too long crash * Second attempt at unbreaking the Travis build * Third attempt at fixing the Travis builds * +tck reactive-streams#308 allow configuring "no events during N time" separately * Update to Gradle 2.12 * Reintegrate dangling footnote in Publisher section. - integrate the footnote in rule 1.9 - sign the Copyright Statement * Asynchronous vs Synchronous Processing: reword "push-based stream" * =tck fixes minor misalignment between code and comment, found via .NET port Semantics remain exacly the same, the error we're testing here is about signaling one more element if request comes in again (which we'll do anyway, regardless of status of this flag) * adjust Subscription.cancel javadoc because cancel command does not have to be called asynchronously * Updating Typesafe to Lightbend * Fix a typo in org.reactivestreams.example.unicast.AsyncSubscriber * Add @seratch to CopyrightWaivers.txt * Fixes reactive-streams#333 by adding license headers to /examples/* * Adds a Glossary, Intent-sections and harmonizes verbiage * Clarifying that object equality is a.equals(b) in Intent for 2.12 * add license header to API directory * add license header to TCK * Fix missing cancel() from in tests that don't consume the entire source * Run with default TestEnvironment settings. * Update CopyrightWaivers.txt * =build reactive-streams#349 equal osgi manifest version as real version To have a tangible PR to talk about. Probably enough to resolve reactive-streams#349 Would be followed up with change to 1.0.1 eventually. * Add Javadoc explanation to the TCK test methods about what they do * Don't import org.reactivestreams.tck.TestEnvironment * Fix missing Javadoc tags * TCK: Request -1 in 309 instead of a random non-positive number * Remove the Random instance as well. * Keep the randomness. * Fixing typos in README.md * Minor rewording of 2.6 to make it easier to understand. (reactive-streams#342) * Minor rewording of 2.6 to make it easier to understand. * Fix spelling errors and clarify a couple of sentences * extra coordination * Remove vague statements, be more specific in others * Update javadoc based on ktoso's feedback * Use the wording eagery for error publisher test 104 * Address feedback, add links to the rules in the javadoc * SubscriberBlackboxVerificationRules explained * Non-BC for TCK: Corrects a typo in test method from *Compuatation to *Computation * Adding a glossary item for external synchronization * Repointing links to sources in README to current main release * =tck reactive-streams#362 signal onComplete in 201 blackbox verification * +tck reactive-streams#362 complete subscriber under test once done in 205 * +tck reactive-streams#362 wait for request signal in 209, and new additional tests * =tck check isCancelled in 205 blackbox; sample the state sometimes * =tck reactive-streams#362 blackbox 209 must issue onSubscribe before any other signal * Clarifies the meaning of "stochastic" for skipStochasticTests() * add additional test for optional_spec111. * now test verifies https://github.com/reactive-streams/reactive-streams-jvm#1.11 and https://github.com/reactive-streams/reactive-streams-jvm#1.5 for publishers, if they support multiple subscribers. * add delegate to IdentityProcessorVerification. * add tests for optional_spec111_registeredSubscribersMustReceiveOnNextOrOnCompleteSignals. * additional happy and the failure cases. * clear typos and change comments. * add new PublisherVerification for multi-subscribers tests. * removed onSubscribe constructor call. renamed Demand -> CancelableSubscription. * Change subscription remove logic. * add myself to CopyrightWaivers. * fix tests by using proxied subscriber, thanks Viktor for helping push this fix * Be consistent in reference style We use the `#.##` style in referring to rules everywhere, this one ref was using a different style - fixed that. * Switching to consistent use of apostrophe in spec * More apostrophe fixes * add patriknw to CopyrightWaivers * Version 1.0.1 * =spec reactive-streams#384 amend spec to allow not mentioning rule number in exception message * Update README.md * =tck reactive-streams#384 dont check for cause message when checking 3.9 * Updating versions to 1.0.1-RC2 and clarifying changes in RELEASE-NOTES.md * Fix links to "Terminal state" (reactive-streams#389) * Fix links to "Terminal state" * add angelsanz to CopyrightWaivers.txt * Preparing 1.0.1 (reactive-streams#390) * Bridge between Reactive-Streams and JDK 9 Flow API (reactive-streams#296) * Bridge between Reactive-Streams and JDK 9 Flow API * Apply changes based on ktoso's feedback * Use oraclejdk9, resolve build.gradle conflict * Change txt/code to use "Reactive Streams" as designator * NPE to use the updated parameter name. * Rename bridge class, tester class (+javadoc) * Java 9 Flow bridge: add Subscriber converters (reactive-streams#399) * Java 9 Flow bridge: add Subscriber converters * Fix return type javadoc * Example synchronous range Publisher (reactive-streams#395) * Example synchronous range Publisher * Udpated with rule numbers in comments * Mentioning rule 3.9 again in emit() * Move classes to the unicast package. * [WIP] TCK for j.u.c.Flow types "directly" (reactive-streams#398) * Add JDK9 TCK, using adapters * Fixing wrapping and unwrapping of the wrappers themselves. * Renames the converters to "toX" for RS and "toFlowX" for Flow. Fixes so that the dist url for gradle is http iso https (TravisCI bug?) Adds regression test for bridge converters. * fix formatting * cleanup

Minor rewording of 2.6 to make it easier to understand.

328c60f

viktorklang added this to the 1.0.1 milestone Mar 5, 2017

viktorklang self-assigned this Mar 5, 2017

Second attempt at rewording 2.6

c3f8b64

viktorklang commented Mar 6, 2017

View reviewed changes

akarnokd approved these changes Apr 4, 2017

View reviewed changes

smaldini approved these changes Apr 11, 2017

View reviewed changes

viktorklang merged commit e884980 into master Apr 12, 2017

viktorklang deleted the wip-321-2.6-rewording-√ branch April 12, 2017 07:24

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Minor rewording of 2.6 to make it easier to understand. #342

Minor rewording of 2.6 to make it easier to understand. #342

viktorklang commented Mar 5, 2017

ktoso commented Mar 5, 2017

akarnokd commented Mar 5, 2017

ktoso commented Mar 5, 2017

viktorklang commented Mar 5, 2017

ktoso commented Mar 5, 2017 •

edited

Loading

viktorklang commented Mar 5, 2017 •

edited

Loading

ktoso commented Mar 5, 2017

viktorklang commented Mar 5, 2017

ktoso commented Mar 5, 2017

viktorklang commented Mar 6, 2017

ktoso commented Mar 6, 2017 •

edited

Loading

viktorklang Mar 6, 2017

viktorklang Mar 6, 2017

viktorklang commented Mar 13, 2017

viktorklang commented Apr 4, 2017

smaldini commented Apr 11, 2017

Minor rewording of 2.6 to make it easier to understand. #342

Minor rewording of 2.6 to make it easier to understand. #342

Conversation

viktorklang commented Mar 5, 2017

ktoso commented Mar 5, 2017

akarnokd commented Mar 5, 2017

ktoso commented Mar 5, 2017

viktorklang commented Mar 5, 2017

ktoso commented Mar 5, 2017 • edited Loading

viktorklang commented Mar 5, 2017 • edited Loading

ktoso commented Mar 5, 2017

viktorklang commented Mar 5, 2017

ktoso commented Mar 5, 2017

viktorklang commented Mar 6, 2017

ktoso commented Mar 6, 2017 • edited Loading

viktorklang Mar 6, 2017

Choose a reason for hiding this comment

viktorklang Mar 6, 2017

Choose a reason for hiding this comment

viktorklang commented Mar 13, 2017

viktorklang commented Apr 4, 2017

smaldini commented Apr 11, 2017

ktoso commented Mar 5, 2017 •

edited

Loading

viktorklang commented Mar 5, 2017 •

edited

Loading

ktoso commented Mar 6, 2017 •

edited

Loading