Thanks Sven... and also thanks very much for your NeoCSV package [1]
[2]. Coincidently my original question was related to practicing how to
use NeoCSV to read some tab-separated data. A bit of feedback on NeoCSV...
1. Very easy to understand and I loved the #recordClass feature
returning a model object.
2. Nicely documented [1]. Also perhaps mention the repository location
and also that it can be found under Pharo Tool > Configuration Browser.
3. My data records had blanks in a lot of fields (eg data1,data2,,data4)
so that #addFloatField: was failing. So I extended with the following
method...
----
addFloatField: accessor ifFail: failBlock
self
addField: accessor
converter: [ :string | Float readFrom: string ifFail: failBlock ]
----
Perhaps you could add similiar, as well as a default 'failBlock' (maybe
one per Integer/Float)
4. Before I added (3.), the debugger would open about eight levels
deeper than required to identify the culprit in the file. It would be
useful if a custom Exception could be generated that held the line and
accessor that was being processed (not that I'm know anything about
Exceptions to know if that would be possible). That might be a good
default 'failBlock'.
[1] https://github.com/svenvc/docs/blob/master/neo/neo-csv-paper.md
[2] http://forum.world.st/ANN-NeoCSV-td4636208.html
[3] http://mc.stfx.eu/Neo
Btw, based on your response to my original post, while I was playing in
Workspace I ended up using it like this....
----change test file source---
csvStream close.
csvStream := FileStream readOnlyFileNamed: 'C:\#ENG8002\Test
Data\Motors\motors.TXT'.
----iteratively-build-and-test-field-list----
csvStream reset.
csvStream nextLine. "ignore headers"
reader := (NeoCSVReader on: csvStream)
separator: Character tab ;
recordClass: LEMotorData .
addIntegerField: #idMotor: ifFail: [ nil ] ;
addFloatField: #hp: ifFail: [ nil ] ;
...etc
reader upToEnd explore.
----etc----
Thanks again.
cheers -ben
Sven Van Caekenberghe wrote:
Ben,
On 03 Nov 2012, at 03:23, Ben Coman <[email protected]> wrote:
I am refereshing my familiarity with streams by looking at PharoByExample1
which says...
"You have to remember that each time you open a stream on a file, you have to close
it too."
Does that mean if in Workspace I iteratively develop some code to process a
stream, so that I am continually running...
myStream := FileStream readOnlyFileNamed: 'C:\test.txt.
...so that 'myStream' is overwritten each time such that I lose reference to
the previous stream, and I am no longer able to send #close to it,
am I creating a memory leak in my image ?
My prior assumption has been that once 'myStream' drops the reference drops to
the previous stream, garbage collection would take care of 'everything'.
cheers -ben
Yes, external resources need to be closed, even though finialization (the
garbage collector telling an object that it is about to become garbage) will
try to clean up as well but it might take some time.
The proper idiom to use is
FileStream readOnlyFileNamed: 'C:\test.txt' do: [ :stream | … ].
or in Pharo 2.0
'C:\test.txt' asFileReference readStreamDo: [ :stream | … ].
'C:\test.txt' asFileReference writeStreamDo: [ :stream | … ].
HTH,
Sven
--
Sven Van Caekenberghe
http://stfx.eu
Smalltalk is the Red Pill
