Hi all,
I just stumbled upon the behavior of #peekBack. I expected that it would give me a "preview" of what I would get when I send #back, but without changing the position of the Stream (like #peek does for #next). Or in other words, I expected to get the last read (or written) element.
As it turned out it answers the next-to-last element instead. There even is a test case in RWBinaryOrTextStreamTest for this behavior.
But does it make sense? I have my doubts mainly because of the inconsistency with the return value of #back.
There is also #peekLast, which does what I expected, but it is only defined for File- and WriteStream and the comment says it is intended to get the last put element, not a last read/extracted/next-ed element.
Kind regards, Jakob
When I read that first line of your post, my impression/expectation was the same as yours. Intuitively #back and #peekBack should return the same value. But my comments are face-value only, without a deep familiarity of streams. Maybe someone else can provide historical & future context.
Presuming it was broadly accepted this "should" be the behaviour of #peekBack, the problem with changing its behaviour is the effect on existing users, and also cross-dialect compatability. "Ideally" Squeak/Cuis would be ammendable to such a change, and then one approach could be a little dance using an interim #peekBackFromCursor so the existing #peekBack can be deprecated for one version (to avoid silent corruption of application logic) then restoring #peekBack with the new behaviour. A slow process, but such small changes accumulate over time to improve things.
Alternatively, we just leave #peekBack as-is and permanently introduce #peekBackFromCursor or similar. Maybe such changes would be superseded by XStreams anyway (not really a reason stop improving current system anyway.)
cheers -ben
On Wed, Jun 21, 2017 at 9:00 PM, Jakob Reschke <jakob.reschke@student.hpi.de
wrote:
Hi all,
I just stumbled upon the behavior of #peekBack. I expected that it would give me a "preview" of what I would get when I send #back, but without changing the position of the Stream (like #peek does for #next). Or in other words, I expected to get the last read (or written) element.
As it turned out it answers the next-to-last element instead. There even is a test case in RWBinaryOrTextStreamTest for this behavior.
But does it make sense? I have my doubts mainly because of the inconsistency with the return value of #back.
There is also #peekLast, which does what I expected, but it is only defined for File- and WriteStream and the comment says it is intended to get the last put element, not a last read/extracted/next-ed element.
Kind regards, Jakob
Hi Jakob, wellcome to the club of peekBack deceived http://lists.squeakfoundation.org/pipermail/squeak-dev/2014-July/179285.html
we can't reasonnably drag too many of these inconsistencies...
2017-06-22 5:48 GMT+02:00 Ben Coman btc@openinworld.com:
When I read that first line of your post, my impression/expectation was the same as yours. Intuitively #back and #peekBack should return the same value. But my comments are face-value only, without a deep familiarity of streams. Maybe someone else can provide historical & future context.
Presuming it was broadly accepted this "should" be the behaviour of #peekBack, the problem with changing its behaviour is the effect on existing users, and also cross-dialect compatability. "Ideally" Squeak/Cuis would be ammendable to such a change, and then one approach could be a little dance using an interim #peekBackFromCursor so the existing #peekBack can be deprecated for one version (to avoid silent corruption of application logic) then restoring #peekBack with the new behaviour. A slow process, but such small changes accumulate over time to improve things.
Alternatively, we just leave #peekBack as-is and permanently introduce #peekBackFromCursor or similar. Maybe such changes would be superseded by XStreams anyway (not really a reason stop improving current system anyway.)
cheers -ben
On Wed, Jun 21, 2017 at 9:00 PM, Jakob Reschke < jakob.reschke@student.hpi.de> wrote:
Hi all,
I just stumbled upon the behavior of #peekBack. I expected that it would give me a "preview" of what I would get when I send #back, but without changing the position of the Stream (like #peek does for #next). Or in other words, I expected to get the last read (or written) element.
As it turned out it answers the next-to-last element instead. There even is a test case in RWBinaryOrTextStreamTest for this behavior.
But does it make sense? I have my doubts mainly because of the inconsistency with the return value of #back.
There is also #peekLast, which does what I expected, but it is only defined for File- and WriteStream and the comment says it is intended to get the last put element, not a last read/extracted/next-ed element.
Kind regards, Jakob
On Fri, Jun 30, 2017 at 5:56 AM, Nicolas Cellier < nicolas.cellier.aka.nice@gmail.com> wrote:
"peekBack deceived"
Strangely, I can't shake the idea this is a good name for a SciFi Thriller movie. cheers -ben
http://lists.squeakfoundation.org/pipermail/squeak-dev/2014- July/179285.html
we can't reasonnably drag too many of these inconsistencies...
2017-06-22 5:48 GMT+02:00 Ben Coman btc@openinworld.com:
When I read that first line of your post, my impression/expectation was the same as yours. Intuitively #back and #peekBack should return the same value. But my comments are face-value only, without a deep familiarity of streams. Maybe someone else can provide historical & future context.
Presuming it was broadly accepted this "should" be the behaviour of #peekBack, the problem with changing its behaviour is the effect on existing users, and also cross-dialect compatability. "Ideally" Squeak/Cuis would be ammendable to such a change, and then one approach could be a little dance using an interim #peekBackFromCursor so the existing #peekBack can be deprecated for one version (to avoid silent corruption of application logic) then restoring #peekBack with the new behaviour. A slow process, but such small changes accumulate over time to improve things.
Alternatively, we just leave #peekBack as-is and permanently introduce #peekBackFromCursor or similar. Maybe such changes would be superseded by XStreams anyway (not really a reason stop improving current system anyway.)
cheers -ben
On Wed, Jun 21, 2017 at 9:00 PM, Jakob Reschke < jakob.reschke@student.hpi.de> wrote:
Hi all,
I just stumbled upon the behavior of #peekBack. I expected that it would give me a "preview" of what I would get when I send #back, but without changing the position of the Stream (like #peek does for #next). Or in other words, I expected to get the last read (or written) element.
As it turned out it answers the next-to-last element instead. There even is a test case in RWBinaryOrTextStreamTest for this behavior.
But does it make sense? I have my doubts mainly because of the inconsistency with the return value of #back.
There is also #peekLast, which does what I expected, but it is only defined for File- and WriteStream and the comment says it is intended to get the last put element, not a last read/extracted/next-ed element.
Kind regards, Jakob
What gives you a view of what would be returned if you sent #back is: PositionableString (and subclasses): #last StandardFileStream (and subclasses): #peekLast * this last one is matches to #back as defined in MutliByteFileStream. #last could be used, but since it uses the internal collection instVar, it may give odd results. I don't know - haven't tried it.
-cbc
On Wed, Jun 21, 2017 at 6:00 AM, Jakob Reschke <jakob.reschke@student.hpi.de
wrote:
Hi all,
I just stumbled upon the behavior of #peekBack. I expected that it would give me a "preview" of what I would get when I send #back, but without changing the position of the Stream (like #peek does for #next). Or in other words, I expected to get the last read (or written) element.
As it turned out it answers the next-to-last element instead. There even is a test case in RWBinaryOrTextStreamTest for this behavior.
But does it make sense? I have my doubts mainly because of the inconsistency with the return value of #back.
There is also #peekLast, which does what I expected, but it is only defined for File- and WriteStream and the comment says it is intended to get the last put element, not a last read/extracted/next-ed element.
Kind regards, Jakob
squeak-dev@lists.squeakfoundation.org
-
Ben Coman -
Chris Cunningham -
Jakob Reschke -
Nicolas Cellier