UTF-8 doc becomes ANSI doc !

gerdb42

Files in ANSI and UTF-8 w/o BOM cannot be distinguished if they do not contain any non-ASCII characters. So the settings Claudia mentioned tell NPP what to assume when it can’t decide upon the encoding (this becomes important once you later add non-ASCII characters).

On the other hand, if a file does contain non-ASCII characters, NPP can clearly detect what the encoding is and it will not perform an automatic conversion.

However, I also don’t get how NPP fits into an automated process. How do you control its actions? Since you know you receive ANSI and need UTF-8 how about throwing together a few lines of C# or powershell?

guy038

Hello, Bahram-Yaghobinia and All,

The encoding problems, generally speaking, coupled with the default Notepad++ behaviour, are difficult enough to understand :-((

In the last part of this post, I give you a possible solution to your problem !?

But, before testing some situations, here are my usual parameters, relative to encoding, of the last N++ version 6.9.2 :

In Settings - Preferences… - New Document - Encoding => Options UTF-8 and Apply to opened ANSI files CHECKED
In Settings - Preferences… - MISC => Option Autodetect character encoding UNCHECKED

In addition, I, generally, have the option Remember current session for next launch, in Settings - Preferences… - Backup, CHECKED

OK. Now, open Notepad++ and let’s make some simple tests :

Open a new file ( CTRL + N ) => Information UTF-8, in the status bar
Type an upper letter A
Save the file, with name Test.txt => Still information UTF-8, in the status bar
Close and restart N++
Click, if necessary, on the Test.txt tab => Still information UTF-8, in the status bar

( An Hex editor would show a one-byte file, of value 41 )

Replace the letter A by the Euro character €, of Unicode value = 20AC ( > \x7F )
Save the changes of Test.txt => Still information UTF-8, in the status bar
Close and restart N++
Click, if necessary, on the Test.txt tab => Still information UTF-8, in the status bar

( An Hex editor would show a three-bytes file, of value e2 82 ac, which is the UTF-8 representation of the Unicode code-point 20AC of the Euro sign )

Now, delete the Euro sign
Save this empty file Test.txt => Still information UTF-8, in the status bar
Close and restart N++
Click, if necessary, on the Test.txt tab => This time, we get, for this empty file, the ANSI information in the status bar !!

( An Hex editor would, effectively, show a zero-byte file )

REMARK :

Although I can understand that N++ cannot decide about the right encoding to choose ( as the file is just empty ) the logic would have been that it chose the default user choice, which is UTF-8 !! Let’s go on :

Type, again, an upper letter A
Re-save the file Test.txt => Still information ANSI, in the status bar
Close and restart N++
Click, if necessary, on the Test.txt tab => We have, again, the information UTF-8, in the status bar

( An Hex editor would show a one-byte file, of value 41 )

For the last time, delete the A character
Save, again, this empty file Test.txt => Still information UTF-8, in the status bar
Close and restart N++
Click, if necessary, on the Test.txt tab => Again, we get, for this empty file, the ANSI information in the status bar !!

( An Hex editor would, effectively, show a zero-byte file )

Type, for the last time, the Euro sign €
Save the changes of Test.txt => Still information ANSI, in the status bar
Close and restart N++
Click, if necessary, on the Test.txt tab => This time, the information remains ANSI, in the status bar !!

( An Hex editor would show a one-byte file, of value 80, as the ANSI-1252 value, for the Euro character, is just … \x80 )

As you, certainly, would like to obtain an UTF-8 encoding, then, you would have to use the menu option Encoding - Convert to UTF-8 and re-save the file Test.txt, because N++ changed it, again, into a three-bytes file, with values e2 82 ac !

Beware not to use the option Encoding - Encode in UTF-8, which only tries to re-interpret the present contents of the file, in the UTF-8 encoding => An unknown one-byte x80 character, which is an invalid UTF-8 value !

So, from above, Bahram-Yaghobinia, it’s obvious that the simple UTF-8 encoding, ( formally named UTF-8 without BOM ) must NOT be used. I, strongly, advice you to adopt the UTF-8 BOM, instead ! Why ?

Just because the invisible BOM ( for Byte Order Mark ) identifies, without any ambiguity the encoding of a file !

When you save, from within Notepad++, a file, with the UTF-8 BOM encoding, three invisible bytes ( EF BB BF ) are added, at the very beginning of your file. These three bytes are, simply, the UTF-8 representation of the Byte Order Mark (BOM) of Unicode code-point FEFF

Thus, each time that N++, ( or any modern editor ! ) will open this file, it automatically understands that it’s a true UTF-8 file, due these invisible three bytes EF BB BF, located at the its beginning !

If you reproduce all the tests above, on a new file, with the UTF-8 BOM encoding ( instead of UTF-8 ), this encoding will remain UTF-8 BOM, throughout all the tests, even when the test file is empty ( just note that, in this specific case, the file is not entirely empty, as it, still, contains the three bytes of the Byte Order Mark !! )

Further information on :

http://en.wikipedia.org/wiki/Byte_order_mark

http://en.wikipedia.org/wiki/Unicode_Specials

http://en.wikipedia.org/wiki/Endianness

Best regards

guy038

Bahram Yaghobinia

Thank you for all the details. I tested your logic and all went well.
I changed my set up to Settings - Preferences… - New Document - Encoding => Options UTF-8 BOM.
Still the same issue but now all my files are saved as Encoding ANSI.
I think the only way to do this is to write a few lines of codes to covert from ANSI to UTF-8.

Claudia Frank

@Bahram-Yaghobinia and all,

maybe I misunderstood the topic but I thought the problem is that a xml file,
which has been created outside from npp gets loaded, manipulated and saved as txt.
Is this the case? If not, why not explaining the steps in detail? Otherwise we
are fishing in the dark to try to find a solution for you.

In addition, i totally agree what guy wrote about detecting utf-8 files by npp,
but unfortunately there is also a reason not to use UTF-8 with BOM.
In the case your manipulated data gets loaded/processed by other applications,
e.g. databases, webserver, web framworks etc… it might be that those apps
can’t handle those data correctly.
Unfortunattely, there are still many of those applications active,
which don’t support UTF-8 BOM files.
I don’t know if this is the case for you - so, just for information.

Cheers
Claudia

Bahram Yaghobinia

• Data is sitting in the queue.
• A job picks up the data. Converts to xml and saves the xml in text file.
• Brings it to the server and saves it on disk
• Right now Notepad++ is the default text editor on the server
• All text files are saved as UFT-8 except the files that have special characters (ALT ####. These files are saved as ANSI).
• Notepad++ set up:
o Setting --> Preference --> New document: I have UTF-8 Apply to opened ANSI files is checked.
o Unchecked the "Auto detect character encoding.
I am trying to add a script that will do this for me, but no idea how it works. Something like (notepad.runMenuCommand(“Encoding”, “Convert to UTF-8”)).

guy038

Hi, Bahram-Yaghobinia,

I was quite surprised and really sorry that your problem isn’t solved at all and seems even worse than before :-(( But, after some minutes, I just realize that this behaviour is quite logical :

As you changed the default encoding, for a new document, to UTF-8 BOM, which hasn’t, of course, an option relative to opened ANSI files, Notepad++ will never try to change an ANSI-style file read, in a true UTF-8 file !
The contents of your file, opened with your default editor N++ 6.9, seem to be, only, one-byte characters. So, as well as characters with value < \x80, the characters, as, for instance, £ © ® — or €, are, also, written with a one-byte sequence, between \x80 and \xFF. Therefore, N++ always saved it, with its present ANSI encoding, without any encoding conversion !

( See the list of all of them, using the N++ menu option Edit - Character Panel, for values > 127 )

So, a solution would be to run a simple script, when starting N++, which :

would apply the menu option Encoding - Convert to UTF-8 BOM
would save the new UTF-8 contents of your file

I think that a Python or NppExec script should do that job, easily !

For information, let’s suppose the exact text £ , © , ® , — or €, in a new file, with an ANSI encoding. This text would produce the sequence of bytes :

A3 20 2C 20 A9 20 2C 20 AE 20 2C 20 97 20 6F 72 20 80 ( 18 bytes )

Once, this text converted with the UTF-8 BOM encoding, it would give the sequence of bytes, below :

EF BB BF C2 A3 20 2C 20 C2 A9 20 2C 20 C2 AE 20 2C 20 E2 80 94 20 6F 72 20 E2 82 AC ( 28 bytes )

I just indicated, in bold, the values of the five characters, which have a different representation in ANSI and UTF-8 BOM encodings, as well and the value of the BOM, at the beginning of the second sequence, in red

Cheers,

guy038

Claudia Frank

This post is deleted!

Claudia Frank

@Bahram-Yaghobinia

I agree with guy, but I would say that your process, which creates the xml needs to take care
about it as utf-8 is the standard encoding for xml. If this process was designed for writing xml
it should have an option to save as utf-8 encoded. Did you double check this?

Cheers
Claudia

Jim Dailey

@Bahram-Yaghobinia

Do the XML files contain something like this as their first line:

<?xml version="1.0" encoding="???" ?>

If so, can you provide that line to us?

If there is no such line, or if it does not include information about the encoding method, then UTF-8 is assumed.

I think that means that if the file contains a £ encoded as “A3” instead of “C2 A3”, that it isn’t technically valid XML (because it isn’t encoded as UTF-8).

Guy, Claudia, or anyone else who has a better understanding please correct me if I am wrong.

Bahram Yaghobinia

Jim, all the XMLs have
<?xml version=“1.0” encoding=“UFT-8” ?>
I am working on the source as well to see if there is anything can be done to save as UFT-8. I would like to approach the script option (Python), but at this time I do not have any idea how it is done.

Jim Dailey

@Bahram-Yaghobinia
Sorry, I can’t help you with Python, but it seems like this part of the process:

A job picks up the data. Converts to xml and saves the xml in text file.

is broken because it does not (always) create valid XML. The XML is invalid any time it claims to be in UTF-8 but includes characters encoded in a single byte (e.g. a “£” encoded as 0xA3) that require multiple bytes to be properly encoded in UTF-8 (e.g. “£” should be encoded as 0xC2 0xA3).

Claudia Frank

@Bahram-Yaghobinia

I still think this is the wrong way to solve the problem because you loose the automatism by interacting
with npp to run the script, but you insist on the python script plugin solution so the lines in question are

notepad.runMenuCommand("Encoding", "Convert to UTF-8")
notepad.save()

But this does only work if you have an english ui, in case you use a different language than you have to replace
“Encoding” and “Convert to UTF-8” with the ones from your language.

Cheers
Claudia

Mahabarata

For me the problem is rather a problem of npp than anything.

The encoding of a text with no special character could be anything : iso-8859-1, iso-8859-2, iso-8859-15, windows-1252, utf-8, ASCII and I imagine a lot of others.
The encoding of a file with only normal characters can’t not be known just by looking at it !!!

In php, there is a function : mb_detect_encoding($txt, array(encoding1, encoding2)).
When you use it, php try to know which encoding is used in the string $txt : it begins to look at the encoding1, if it fits the function answers encoding1, if it doesn’t the function try the second encoding and so and on.

So if there is no special character in the $txt, the function will tell you the encoding is encoding1. In your case, with encoding1=utf-8, the function will tell you that the $txt is a utf-8 even if it’s impossible to know !

npp has no option to do the same (it is what I was looking for between my first post and my second one) except the “Settings - Preferences… - New Document - Apply to opened ANSI file” (thanks again Claudia).
You can’t obtain from npp that your docs are forever iso-5988-1, iso-5988-2 or ASCII ! For me it’s a big pb because a polish guy will probably use an ISO-5988-2 encoding and npp will tell some docs are in ANSI (that is windows-1252) and others are in ISO-5988-2 !

So the pb is not only a pb with utf-8/ANSI but with a lot of encodings !

I think it will be a good evolution of npp to add an option to tell what to do when a doc is open and it is impossible to know the encoding : only the user can tell it, npp what clever it is will never !

Claudia Frank

@Mahabarata

you are correct, there is no way to always guess the correct encoding.
In regards to phps mb_detect_encoding function, npp is,
when “Autodetect character encoding” is checked, using mozillas chardet library,
so it has such functionality but, as you found already out, cannot guess the
correct encoding all time.

I would also find it very useful if the setting
New Document->Encoding: UTF-8 and Apply to opened ANSI files (or any other configured encoding)
would force npp to treat all new opened documents as “configured encoding” when
auto detection of encoding has been disabled.

Cheers
Claudia

gerdb42

@Claudia-Frank

Let’s assume a file contains Byte-sequence 20-A9-20 (in ANSI this would be Space-Copyright-Space). This Sequence is invalid in UTF-8 so NPP has no alternative other than assuming an single-Byte encoding. And since it never does changes to the file’s content on its own, it is left to treat such a file as ANSI (or whatever your favorite single-Byte encoding is).

This is not a shortcoming of NPP but part of that single-Byte heritage we still have to deal with today.

Claudia Frank

@gerdb42

I assume we have the same understanding so I’m interested to know
what I have written that could be misunderstood?
Could you point me to my error?

Thank you and cheers
Claudia

gerdb42

@Claudia-Frank said:
Not quite an error, but

I would also find it very useful if the setting
New Document->Encoding: UTF-8 and Apply to opened ANSI files (or any other configured encoding)
would force npp to treat all new opened documents as “configured encoding” when
auto detection of encoding has been disabled.

would require an implicit conversion to UTF-8. And besides breaking the principle of not doing changes without user action, it will pop up a whole bunch of other issues.

Claudia Frank

@gerdb42

I agree that this would break the principle but on the other hand it could be beneficial as well.
But, now as I’m typing I’m thinking, when this conversion takes place and you don’t know from which encoding it came from
you might corrupt the document without knowing how to fix it.
Yes - bad idea.

Cheers
Claudia