Hey thanks guys - so looking at readStreamEncoded: - how do I know what the valid encodings are? Skimming those doc��s Sven referenced, I can start to pick out some - but is there a list? I see that method parameter says ��anEncoding�� but the type hint on that is misleading as it seems like its a String or is it a Symbol? If I search for Encoder classes - I do find ZnCharacterEncoder - and it has class methods for latin1, utf8, ascii - so is this the definitive list? And should the encoding strings used in those methods be constants or something I can reference in my code?

Gosh - this raises a whole host of things I just naively assumed happened for me.

So it looks like the file giving me issues - seems to have characters like �� or �� in it. So I��m wondering how I know what the proper encoding format would be (I think these files were written out with some PHP app) - is it just a trial and error thing?

I tried changing my code to:

details parseStream: (firmEfs readStreamEncoded: 'iso-8859-1��). - and other variants like ��ASCII�� and ��latin1�� - and this then gives me another error:

"ZnCharacterEncodingError: Character Unicode code point outside encoder range��

So it does sound like I have a file that isn��t conforming to known standards - and I guess I have to use #beLenient option.

Sven - In the examples for using #beLenient - you seem to show something that assumes you will iterate with Do - as my existing code takes a stream, that it wants to do a #nextLine on - would it be bad to do something like this:

efsStream := (firmEfs readStreamEncoded: 'latin1').

efsStream encoder beLenient.

details parsStream: efsStream.

That is - get the endcoder from my Stream and make it lenient?

Appreciate the pointers on this guys - I��m definitely learning something new here.

Tim

On 20 Jul 2021, at 12:11, Guillermo Polito <guillermopolito@gmail.com> wrote:

El 20 jul 2021, a las 11:45, Sven Van Caekenberghe <sven@stfx.eu> escribi��:

On 20 Jul 2021, at 11:03, Sven Van Caekenberghe <sven@stfx.eu> wrote:

Hi Tim,

An introduction to this part of the system is in https://ci.inria.fr/pharo-contribution/job/EnterprisePharoBook/lastSuccessfulBuild/artifact/book-result/Zinc-Encoding-Meta/Zinc-Encoding-Meta.html [Character Encoding and Resource Meta Description] from the "Enterprise Pharo" book.

The error means that a file that you try to read as UTF-8 does contain things that are invalid with respect to the UTF-8 standard.

Are you sure the file is in UTF-8, maybe it is in ASCII, Latin-1 or something else ?

It is possible to customise the encoding to something different than the default UTF-8. For non-UTF encoders, there is a strict/lenient option to disallow/allow illegal stuff (but then you will get these in your strings).

I can show you how to do that if you want.

'/var/log/system.log' asFileReference readStreamDo: [ :in | in upToEnd ].

'/var/log/system.log' asFileReference binaryReadStreamDo: [ :in |
(ZnCharacterReadStream on: in encoding: #ascii) upToEnd ].

'/var/log/system.log' asFileReference binaryReadStreamDo: [ :in |
(ZnCharacterReadStream on: in encoding: ZnCharacterEncoder ascii beLenient) upToEnd ].

There is also readStreamEncoded:[do:], which is a bit more concise but does the same :)

HTH

Sven

On 20 Jul 2021, at 10:31, Tim Mackinnon <tim@testit.works> wrote:

Hi - I��m doing a bit of log file processing with Pharo - and I��ve hit an unexpected error and am wondering what the best way to approach it is.

It seems that I have a log file that has unexpected characters, and so my readStream loop that reads lines gets an error: "ZnInvalidUTF8: Illegal continuation byte for utf-8 encoding��.

For some reason this file (unlike my others) seems to contain characters that it shouldn��t - but what is the best way for me to continue processing? Should I be opening my files in a different way - or can I resume the error somehow- I��m not familiar with this area of Pharo and am after a bit of advice.

My code is like this (and I get the error when doing nextLine)

parseStream: aFileStream with: aBlock
| line items |
[ (line := aFileStream nextLine) isNil ]
whileFalse: [
items := $/ split: line.
items size = 3 ifTrue: [aBlock value: items]]

My stream is created like this:

firmEfs := (pathName , '/' , firmName , '_files') asFileReference.
details parseStream: firmEfs readStream.

Should I be opening the stream a bit differently - or can I catch that encoding error and resume it with some safe character?

Thanks for any help.

Tim