when I run deploy and assets are precompiled:

bundle exec rake RAILS_ENV=production RAILS_GROUPS=assets assets:precompile"

I get this error:

.../gems/get_pomo-0.6.3/lib/get_pomo/po_file.rb:25: invalid multibyte character: /\xEF\xBB\xBF/

Escaping slashes has some weird behaviour right now.

Imho it should escape slashes if these slashes were used to escape content in the original string. That means

 \" should become \\\"
but
\n should escape to \\n

It somewhat seems to double escape slashes in the comments, but doesn't in the msgstr.

Example:

#, fuzzy
#| msgid "Weird Bug\" :"
msgid "Weirder Bug"
msgstr "Komischer Bug\" :"

(with fuzzy msgstr "Komischer Bug" :" obviously)
Will output to:

Komischer Bug" :       # msgstr
Weird Bug\" :          # fuzzy msgid (#| msgid)

So imho comment output works correctly, but msgstr (and probably msgid) are not escaping the slash before the quote and are therefor generating invalid output if escaped characters like quotes are in the original string.

I have just discovered that this project is no longer maintained. I'll therefore work to migrate to the recommended system.

However, for the record, when there are two identical msgids, one with a context and the other without, the one without context is ignored.

It is unique_translations that drops the missing translations.

mk = ->(ctx, id, str) {
    res = GetPomo::Translation.new
    res.msgctxt = ctx if ctx
    res.msgid = id
    res.msgstr = str
    res
}

translations = [
    mk.(nil, 'Amount', 'Montant'),
    mk.('Invoice', 'Amount', 'Montant'),
]
GetPomo.unique_translations(translations)

=>

[#<GetPomo::Translation:0x00007ffb57cac218
  @msgctxt="Invoice",
  @msgid="Amount",
  @msgstr="Montant">]

If the .po file start with a comment (a #) it raise an error.

ignores obsolete messages (#~) as they do not have msgstr and msgid and are just comments.

message strings are like:
index.rb\n fuzzy\n

so FUZZY_REGEX = /^\s_fuzzy\s_$/

does not match it. Perhaps change to:
/^fuzzy\n$/

The parser does not remember linebreakes of multiline msgid's and msgstr's.

well this is a rather big one I didn't really thought about changing, as I use my own methods for writing the method and not to_text. In the process of refactoring I would like to do anything that is about parsing/writing with the gem.

It does not violate the gettext rules to concatenate multiline msgid's or msgstr's, but it causes for example the header to be in one line, which is not desireable.

To keep the multiline presentation while parsing, msgstr and msgid must be represented as a list (array) internally.
I could do this the harsh way which would require a version jump and breaking any existing use of the gem, or the soft way by adding helper methods for msgid/msgstr which return the concatenated version, but keeping lists for their content.

On a plus the second way wouldn't change how the mo/po interface works and wouldn't break existing code. The drawback would be performance and messy logic.

What do you think @grosser ?

@grosser just wanted to let you know that I worte a faster parser with more detailed output and just open sourced it today https://github.com/experteer/simple_po_parser

It's actually meant as the new core for this project https://github.com/arashm/PoParser

It benchmarks about twice as fast as this parser and has a more detailed output. (It basically splits the output into a hash where every little fragment specified by the GNU gettext specification has it's own key. My private project using the get_pomo gem since 2013 actually parsed this stuff in a second parsing step from the get_pomo output 😃)

I'm not sure if it makes sense to update this old project to use the other projects as new cores and add the unique features of this project as another layer on top, but at least I wanted to let you know it's out there :D

I've a PO file with:

msgctxt "Not
really."
msgid "I have\n%d cat."
msgstr ""

This is raining an GetPomo::InvalidString error.

I'm not sure if this is a valid po format, or each line should end in "