February 2017 - Pharo-dev - lists.pharo.org

Re: [Pharo-dev] WeakValueDictionary: shouldn't be "absent" after GC?

by Guillermo Polito

On Tue, Feb 21, 2017 at 10:37 AM, Denis Kudriashov <dionisiydk(a)gmail.com> wrote: > Hi. > > 2017-02-21 9:44 GMT+01:00 Guillermo Polito <guillermopolito(a)gmail.com>: > >> Hi Dr. Dias, >> >> a WeakValueDictionary get's its values collected if they are not >> referenced strongly. However, the association remains there. This means >> that when you do: >> >> dictionary := WeakValueDictionary with: 'hello' -> nil copy. >> >> You'll have something like this: >> >> WeakValueDictionary { >> WeakValueAssociation { key: 'hello', value: yourObject } >> } >> >> Once you garbage collect, the value is holding yourObject weakly, so >> you'll have: >> >> WeakValueDictionary { >> WeakValueAssociation { key: 'hello', value: nil } >> } >> >> But the WeakValueAssociation will still be there. >> >> > For me what you describe is implementation details and users should not > think about it at all. > Of course it is an implementation detail. I'm just describing how it works (and thus, how it does not work). I never said it was a good or bad approach. I agree ifAbsent~ messages should work with collected values by "absent > branch". It is intuitive. > > To support it correctly we can use approach of Set which uses > #asSetElement to store items. Then we can easily distinguish nil from > absent value. Real nil value will happen only by garbage collection. > But I am wondering WeakSet not uses SetElement approach directly and > instead it uses special flag object. Maybe it should be fixed too. > But in any case, whatever the alternative you give users, they must understand the tradeoff. If all Weak*Dictionaries were subscribed by default to the weak registry for cleanup, that would provoke a lot of overhead during garbage collection. Magic is good in some cases, but when you're dealing with low level stuff explicitly (like instantiating a Weak*Dictionary) I prefer to control some of that magic explicitly.

Feb. 21, 2017

Re: [Pharo-dev] Catalog loading in Spotter

by Torsten Bergmann

As there were no further objections I guess we should do the following steps as a compromise: 1. We switch back to how it was by selecting Solution A as it is the old behavior and the pragmatic approach. This avoids the confusion, allows to quickly search for the catalog items and also renders the videos valid again. 2. Pharo 6 is still not released. People can still be testing and if this really gives trouble again (maybe next time in a reproducible way) we can try to fix it/find another solution. In the worst case we can disable it by disabeling the Catalog setting (not the GT extension itself) so it is not confusing or hidden again in the Pharo settings. I've opened issue https://pharo.fogbugz.com/f/cases/edit/19732 and will fix it. Lets see how it goes. Thanks T. > Gesendet: Montag, 20. Februar 2017 um 11:27 Uhr > Von: "Sven Van Caekenberghe" <sven(a)stfx.eu> > An: "Pharo Development List" <pharo-dev(a)lists.pharo.org> > Betreff: Re: [Pharo-dev] Catalog loading in Spotter > > > > On 20 Feb 2017, at 10:30, Pavel Krivanek <pavel.krivanek(a)gmail.com> wrote: > > > > > > > > 2017-02-20 10:27 GMT+01:00 Guillermo Polito <guillermopolito(a)gmail.com>: > > As far as I remember, the main problem was not bandwidth but a blocking and non-responsive UI. > > Imagine I'm in the university, or in a company with a proxy. And I forget to set the proxy. Then, while spotter tries to fetch catalog entries, it would block the image and throw a timeout error some minutes after. > > > > So, maybe for the future and besides Ben's C option (which I like) I have a > > > > D) Load catalog entries in background in a non-blocking way > > - and while catalog entries are not yet available, they should not be shown > > > > AFIK the main problem was that there was not possilbe to do it non-blocking way. > > Yes that is true. (It is actually DNS resolving that is blocked in some freak case). > > But the thing is, the problematic situation only occurs for a very small percentage of people, like less than 1%, probably even much less. (Remember, nobody can produce a repeatable case for it). > > So the question is, is that enough reason to act as if we are still in the 90's when internet was a luxury and cut features that depend on it ? If you have no network today, you're in big trouble anyway. (Again, if you turn off your network, no blocking will happen at all, try it). > > So what is more important ? The general functionality of the 99.xxx % or the unreproducible freak case ? > > > -- Pavel > > > > > > Even further, It would make sense a solution that mixes Ben's caching with background loading of the catalog entries... > > > > Guille > > > > On Mon, Feb 20, 2017 at 1:03 AM, Ben Coman <btc(a)openinworld.com> wrote: > > On Mon, Feb 20, 2017 at 4:05 AM, Torsten Bergmann <astares(a)gmx.de> wrote: > > > In the past in Pharo it was possible to open Spotter, type in the name of a framework/project to load > > > from catalog, perform a search and just hit ENTER to easily install the project. > > > > > > This was following the Spotter idea that it is easy to access most informations of Pharo > > > with the Spotter tool. > > > > > > There always was and still is a setting in "Settings" -> "Catalog" -> "Display catalog projects in Spotter". > > > This setting is ENABLED BY DEFAULT but could be switched off in the settings tool or custom preference > > > scripts if this is problematic for someone. > > > > > > > > > Now in Pharo 6 there is an additional class "GTSpotterExtensionSettings" to activate/deactivate > > > Spotter extensions. While nearly all of the Spotter extensions are enabled the one for the catalog > > > integration is DISABLED BY DEFAULT. > > > > > > This leads to several effects: > > > > > > 1. While in the past it was possible in a fresh Pharo image to search and install out of > > > the box it is (as of today in the not yet released Pharo 6) not possible anymore to quickly > > > start by searching and installing from catalog using Spotter. > > > > > > 2. It is very confusing that in the settings "Display catalog projects in Spotter" is enabled but a search > > > in Spotter gives no results. Most people will not not know about the second setting and easily > > > get lost and think this behavior is just broken. > > > Also this second setting for the Spotter extension is much more hidden between all the other > > > Spotter extension enablements very and hard to find. > > > > Generally its not good to have two settings for the one thing. > > > > > > > > 3. Several of my youtube videos demonstrating Goodies like DesktopManager, QuickAccess, > > > MessageFlowBrowser, ... directly start by loading the tools from Spotter. Anyone newbee who will > > > follow these will not only be confused - but also stuck in trying Pharo when he learns > > > from these videos. > > > > UI things do evolve and videos date. But these seem worthwhile to keep current. > > > > > > > > I was asked several times on Slack and via Mail from people who were not able to reproduce ... this > > > is really annyoing. Especially this gives the wrong impression to newbees. Things should be easy > > > not complicated. > > > > > > To my knowledge disabling the Spotter search in Pharo 6 came up due to some Pharo teaching in regions > > > with slow internet connection. > > > > This made sense at the time. It is a worse first impression if UI > > lags when this central part of the UI is interfaced. > > How often would it access the network? Every time? Or did it cache > > the result for a day, etc? > > > > > I understand that we would like to support these Pharo users too as best > > > as possible in their out of the box experience ... but (without being able to prove) I think that 90% or > > > more Pharo users have a regular internet connection. Otherwise it would be hard to work with updates, > > > project loading, PharoLauncher, STHub or Iceberg/GitHub. > > > > > > Also my own personal experience is that even on low bandwidth network this Catalog Spotter search for > > > me was always fast enough (as I often use Pharo in trains with slow connections or on a Pi with slow > > > connections and less processing power). I do not know about all others from the community. > > > > > > I invested hours in the past in developing and introducing the initial configuration browser to Pharo, > > > later improved and helped shaping its replacement CatalogBrowser, also contributed this spotter search > > > for the catalog items so things are more accessible, easy and enjoyable. That's why I also invested > > > hours in udpating configs or pushing you to put things into catalog. > > > > > > Because accessibility is key. Only when things are easy to access and understandable people will > > > enjoy Pharo. > > > > > > Currently in an out-of the box image this easy access to the projects via Spotter is blocked. > > > Additionally I have to explain to anyone who asks me that there is a second non-obvious/more hidden setting > > > leaving an unpleasant feeling how many others unknown to me will struggle with this issue. > > > > > > I see two solutions: > > > > > > A) We enable both settings by DEFAULT to bring back the Spotter search and installation > > > of catalog items - with the clear benefit of having > > > - the previous behavior in Pharo back > > > - the out of the box ability to search for catalog projects in any fresh image > > > - no confusion among the user base anymore regarding the settings > > > - we have unbroken Youtube videos that newbees can continue to follow > > > - if a user asks (like often) how to get Seaside, Artefact, Mongo, Teapot or other projects we can > > > just tell him "search in Spotter and you should be fine" as most of them have a config in > > > the catalog. > > > > > > Remember that not all of us know about all the github pages or nice Metacello expressions. > > > So the easier things are found and accessible the better it is. > > > > > > B) If A is still a "No go" for the community we should at a minimum switch the defaults of > > > the two settings: > > > > > > => we ENABLE the Spotter extension (GTSpotterExtensionSettings perform: #GTSpotter_spotterCatalogProjectsFor: with: true) > > > => we DISABLE the catalog setting (CatalogSettings displayCatalogProjectsInSpotter: false) > > > > > > With this at least we have no confusion among the user base anymore regarding the settings. > > > > C) Could a STON file of the catalog be cached in the zip of > > Image/Changes files? > > That zip might be updated daily to keep it current. > > Then in a fresh image Spotter searches the cached file for a instant response, > > while the Catalog is updated from the network in the background. > > > > Even with fast Internet the Catalog should only need to auto-update once a day, > > with results cached in a common file to be accessed when fresh images > > are started. > > > > The Spotter UI might display the last update date as part of the > > Catalog sub-heading in its results. > > > > cheers -ben > > > > > > > > I would clearly and strongly vote for option A as my preferred one. > > > > > > I agree there are regions/continents with very low bandwidth - but Pharo will rival with state of the art > > > technologies where loading/installation megabytes from the web is often not seen as an issue. There are > > > many package registries out there (from debian packages) up to Maven, npm in JS, ... or look at Docker. > > > Shuffling megabytes around is a reality in todays technologies. > > > > > > So to be honest I never understood this whole "bandwidth" discussion and even if this comes up it could > > > be solved with a note in the download/welcome screen or pointing to a custom preference script for low bandwidth > > > situations. > > > > > > Sorry for having to bring this up again ... but I would like this to be solved BEFORE Pharo 6 will > > > be pushed out of the door. Keeping it like it is without further actions would be really stupid. > > > > > > Thanks for you comments, ideas or votes. > > > > > > Thanks > > > T. > > > > > > > > > > > >

Feb. 21, 2017

Re: [Pharo-dev] [pharo-project/pharo-core] 986ba2: 60402

by Andrei Chis

Ping me in the evening when you have time. I might have some time but I'm not 100% sure. Cheers, Andrei On Tue, Feb 21, 2017 at 7:47 AM, phil(a)highoctane.be <phil(a)highoctane.be> wrote: > There are quite a few Themers + hard coded colors in GT. > > Maybe someone from the GT team can pair for an hour or two with me so that > we can get this done. > > I can do this in the evening CET as I am pretty busy this week (managed to > have Pharo a large part of my daily work now). > > Phil > > On Mon, Feb 20, 2017 at 6:54 PM, Sven Van Caekenberghe <sven(a)stfx.eu> > wrote: > >> Thanks. >> >> Not that it is a big deal, but the default background does not really >> match, but that might be taste as well. >> >> > On 20 Feb 2017, at 17:41, phil(a)highoctane.be wrote: >> > >> > I have fixed the spotter styling but need to get a 6.0 and move it >> there. >> > >> > <image.png> >> > >> > On Mon, Feb 20, 2017 at 5:25 PM, Sven Van Caekenberghe <sven(a)stfx.eu> >> wrote: >> > >> >> On 20 Feb 2017, at 16:45, GitHub <noreply(a)github.com> wrote: >> >> >> >> 19728 Integrate Sublimish theme >> >> https://pharo.fogbugz.com/f/cases/19728 >> > >> > Very nice ! >> > >> > But does it also not need Spotter styling ? This looks odd: >> > >> > <Screen Shot 2017-02-20 at 17.23.27.png> >> > >> >> >> >

Feb. 21, 2017

Re: [Pharo-dev] WeakValueDictionary: shouldn't be "absent" after GC?

by Denis Kudriashov

Hi. 2017-02-21 9:44 GMT+01:00 Guillermo Polito <guillermopolito(a)gmail.com>: > Hi Dr. Dias, > > a WeakValueDictionary get's its values collected if they are not > referenced strongly. However, the association remains there. This means > that when you do: > > dictionary := WeakValueDictionary with: 'hello' -> nil copy. > > You'll have something like this: > > WeakValueDictionary { > WeakValueAssociation { key: 'hello', value: yourObject } > } > > Once you garbage collect, the value is holding yourObject weakly, so > you'll have: > > WeakValueDictionary { > WeakValueAssociation { key: 'hello', value: nil } > } > > But the WeakValueAssociation will still be there. > > For me what you describe is implementation details and users should not think about it at all. I agree ifAbsent~ messages should work with collected values by "absent branch". It is intuitive. To support it correctly we can use approach of Set which uses #asSetElement to store items. Then we can easily distinguish nil from absent value. Real nil value will happen only by garbage collection. But I am wondering WeakSet not uses SetElement approach directly and instead it uses special flag object. Maybe it should be fixed too.

Feb. 21, 2017

Re: [Pharo-dev] WeakValueDictionary: shouldn't be "absent" after GC?

by Guillermo Polito

Hi Dr. Dias, a WeakValueDictionary get's its values collected if they are not referenced strongly. However, the association remains there. This means that when you do: dictionary := WeakValueDictionary with: 'hello' -> nil copy. You'll have something like this: WeakValueDictionary { WeakValueAssociation { key: 'hello', value: yourObject } } Once you garbage collect, the value is holding yourObject weakly, so you'll have: WeakValueDictionary { WeakValueAssociation { key: 'hello', value: nil } } But the WeakValueAssociation will still be there. Apparently, what you want to do is to automatically clean your dictionary. Checking a Pharo60371, I see that WeakValueDictionaries do not implement this... But WeakKeyDictionary does. Take for example a look at WeakKeyDictionary>>finalizeValues. If you subscribe the WeakKeyDictionary instance to the WeakRegistry, the weak registry will send finalize values to your weak key dictionary and they will eliminate the expired associations. However, #finalizeValues is not implemented in WeakValueDictionary. Maybe we should? Or maybe you should try with Ephemerons. Guille BTW why are you using a copy of nil? you can test it with a plain new Object (i.e., Object new). It took me a couple of extra seconds to realize what you were doing. On Tue, Feb 21, 2017 at 7:57 AM, phil(a)highoctane.be <phil(a)highoctane.be> wrote: > This makes sense, no? > > | dictionary | > dictionary := WeakValueDictionary with: 'hello' -> nil copy. > "Smalltalk garbageCollect." > r:={ > dictionary values includes: nil. > dictionary at: 'hello'. > dictionary at: 'hello' ifAbsent: [ 'absent' ]. > dictionary at: 'hello' ifAbsentPut: [ 'put' ]. > }. > r > > and I get this anyway when GCing and puttting #x as value > > [image: Inline image 1] > > Maybe checking for emptiness first is needed before going for values > include: > > Phil > > > On Mon, Feb 20, 2017 at 11:18 PM, Martin Dias <tinchodias(a)gmail.com> > wrote: > >> Hi. The answer of: >> >> | dictionary | >> dictionary := WeakValueDictionary with: 'hello' -> 'world' copy. >> Smalltalk garbageCollect. >> { >> dictionary values includes: nil. >> dictionary at: 'hello'. >> dictionary at: 'hello' ifAbsent: [ 'absent' ]. >> dictionary at: 'hello' ifAbsentPut: [ 'put' ]. >> } >> >> is: >> >> #(true nil nil nil) >> >> >> Bug or feature? shouldn't the value be considered as absent instead of >> present (and nil). >> >> cheers, >> MartÃn >> > >

Feb. 21, 2017

Re: [Pharo-dev] WorkingSession UUID looks "sketchy"

by Yuriy Tymchuk

I think that Peterâs main point was that a session should be reinitialized on boot and not reset during the closing event. Uko > On 20 Feb 2017, at 15:51, Sean P. DeNigris <sean(a)clipperadams.com> wrote: > > Guillermo Polito wrote >>> So if image crashes or I am running it headlessly without saving I am >>> actually still on the same session. >>> >> Peter, how can we make this an actionable point? > > Yes, that doesn't sound good! > > > > ----- > Cheers, > Sean > -- > View this message in context: http://forum.world.st/WorkingSession-UUID-looks-sketchy-tp4933140p4935079.h… > Sent from the Pharo Smalltalk Developers mailing list archive at Nabble.com. >

Feb. 21, 2017

Re: [Pharo-dev] [Pharo6][Issue Tracker] Please check Issues tagged Milestone-6.0

by denker

> On 20 Feb 2017, at 08:30, denker <marcus.denker(a)inria.fr> wrote: > > Hi, > > We have right now 112 issues tagged to be fixed before we can release Pharo6: > > https://pharo.fogbugz.com/f/filters/1192/6-0-All <https://pharo.fogbugz.com/f/filters/1192/6-0-All> > > It would be nice if everyone could check the issues and help deciding: > > âis this a blocking issue that holds up release?â > > If no â> remove Pharo6 milestone. Do not put it on Pharo7, just âLaterâ. > We are down to 98 tagged for Pharo6 and 610 open issues over all. Marcus

Feb. 21, 2017

Re: [Pharo-dev] WeakValueDictionary: shouldn't be "absent" after GC?

by phil＠highoctane.be

This makes sense, no? | dictionary | dictionary := WeakValueDictionary with: 'hello' -> nil copy. "Smalltalk garbageCollect." r:={ dictionary values includes: nil. dictionary at: 'hello'. dictionary at: 'hello' ifAbsent: [ 'absent' ]. dictionary at: 'hello' ifAbsentPut: [ 'put' ]. }. r and I get this anyway when GCing and puttting #x as value [image: Inline image 1] Maybe checking for emptiness first is needed before going for values include: Phil On Mon, Feb 20, 2017 at 11:18 PM, Martin Dias <tinchodias(a)gmail.com> wrote: > Hi. The answer of: > > | dictionary | > dictionary := WeakValueDictionary with: 'hello' -> 'world' copy. > Smalltalk garbageCollect. > { > dictionary values includes: nil. > dictionary at: 'hello'. > dictionary at: 'hello' ifAbsent: [ 'absent' ]. > dictionary at: 'hello' ifAbsentPut: [ 'put' ]. > } > > is: > > #(true nil nil nil) > > > Bug or feature? shouldn't the value be considered as absent instead of > present (and nil). > > cheers, > MartÃn >

Feb. 21, 2017

Re: [Pharo-dev] [pharo-project/pharo-core] 986ba2: 60402

by phil＠highoctane.be

There are quite a few Themers + hard coded colors in GT. Maybe someone from the GT team can pair for an hour or two with me so that we can get this done. I can do this in the evening CET as I am pretty busy this week (managed to have Pharo a large part of my daily work now). Phil On Mon, Feb 20, 2017 at 6:54 PM, Sven Van Caekenberghe <sven(a)stfx.eu> wrote: > Thanks. > > Not that it is a big deal, but the default background does not really > match, but that might be taste as well. > > > On 20 Feb 2017, at 17:41, phil(a)highoctane.be wrote: > > > > I have fixed the spotter styling but need to get a 6.0 and move it there. > > > > <image.png> > > > > On Mon, Feb 20, 2017 at 5:25 PM, Sven Van Caekenberghe <sven(a)stfx.eu> > wrote: > > > >> On 20 Feb 2017, at 16:45, GitHub <noreply(a)github.com> wrote: > >> > >> 19728 Integrate Sublimish theme > >> https://pharo.fogbugz.com/f/cases/19728 > > > > Very nice ! > > > > But does it also not need Spotter styling ? This looks odd: > > > > <Screen Shot 2017-02-20 at 17.23.27.png> > > > > >

Feb. 21, 2017

WeakValueDictionary: shouldn't be "absent" after GC?

by Martin Dias

Hi. The answer of: | dictionary | dictionary := WeakValueDictionary with: 'hello' -> 'world' copy. Smalltalk garbageCollect. { dictionary values includes: nil. dictionary at: 'hello'. dictionary at: 'hello' ifAbsent: [ 'absent' ]. dictionary at: 'hello' ifAbsentPut: [ 'put' ]. } is: #(true nil nil nil) Bug or feature? shouldn't the value be considered as absent instead of present (and nil). cheers, MartÃn

Feb. 20, 2017