I just got done reading the Django docs on internationalization and something is bugging me.
It seems like the default language strings are themselves used as the keys for retrieving translations. That's perfectly fine for short text, but for paragraphs, it seems like poor design. Now, whenever I change the English (default) text, my keys change and I need to manually update my .po file?
It seems like using keys like "INTRO_TEXT" and retrieving the default language from it's own .po file is the right approach. How have others approached this problem and what has worked well for you?
Yes, you will have to manually update the PO files, but most of the times it will be limited to remove the fuzzy marks on out-of-date translations (gettext marks translations as fuzzy when the original version is modified).
I personally prefer this approach as it keeps the text content in the source code which makes it much more readable (especially for HTML). I also find it hard to find good and concise string identifiers, poorly named identifiers are headache prone.
Django-rosetta will be of great help if you do not want to edit PO files by hand or want to delegate translations to non developers.
Related
I have a C++ project in which comments of source code are in Chinese language, now I want to convert them into English.
I tried to solve using google translator but got an Issue: Whole CPP files or header didn't get converted, also I have found the name of the struct, class etc gets changed. Sometimes code also gets modified.
Note: Each .cpp or .h file is less than 1000 lines of code.But there are multiple C++ projects each having around 10 files. Thus I have around 50 files for which I need to translate Chinese text to English.
Well, what did you expect? Google Translate doesn't know what a CPP file is and how to treat it. You'll have to write your own program that extracts comments from them (not that hard), runs just those through Google Translate, and then puts them back in.
Mind you, if there is commented out code, or the comments reference variable names, those will get translated too. Detecting and handling these cases is a lot harder already.
Extracting comments is a lexical issue, and mostly a quite simple one.
In a few hours, you could write (e.g. with flex) some simple command line program extracting them. And a good editor (such as GNU emacs) could even be configured to run that filter on selected code chunks.
(handling a few corner cases, such as raw string literals, might be slightly more difficult, but these don't happen often and you might handle them manually)
BTW, if you are assigned to work on that code, you'll need to understand it, and that takes much more time than copy&pasting or editing each comments manually.
At last, I am not sure of the quality of automatic translation of code comments. You might be disappointed. Also, the code names (of functions, of classes, of variables, etc...) matter a lot more.
Perhaps adding your comments in English could be wiser.
Don't forget to use some version control system. You really need one (e.g. git)
(I am not convinced that extracting comments for automatic translation would help your work)
First separate both comment and code part in different file using python script as below,
import sys
file=sys.argv[1]
f=open(file,"r")
lines=f.readlines()
f.close()
comment=open("comment.txt","w+")
code=open("code.txt","w+")
for l in lines:
if "//" in l:
comment.write(l)
code.write("\n")
else:
code.write(l)
comment.write("\n")
comment.close()
code.close()
Now translate comment.txt with google translator and then use
paste code.txt comment_en > source
where comment_en is translated comment in english.
I have a Django Project where I used Sphinx to create my documentation. I went through sphinx-apidoc and ran 'make latexpdf'. The resulting documentation has a quite a few lines that flow out of the margin. On top of margin issues, lines in the index start overflowing onto each other.
Overflowing Lines
Margin Issues :(
Is there an easy way to fix these issues (or an easier way to create PDF documentation)?
ELI5 if possible (I'm not well-versed in LaTeX)
The overflowing lines situation in the index should improve from adding this to conf.py:
latex_elements = {
'printindex': '\\footnotesize\\raggedright\\printindex',
}
Or, you can switch to Japanese language which does something like that (even better) out-of-the box from its special document class ;-)
TeX does not always know how by itself how to insert linebreaks: after all it is good at hyphenation of natural language. But as pointed out in comments Sphinx coerces LaTeX into handling better long code lines since 1.4.2.
Since recent 1.5.3, user can customize page margins, check http://www.sphinx-doc.org/en/stable/latex.html#the-sphinx-latex-style-package-options for documentation of hmargin and vmargin which can be configured via 'sphinxsetup'.
Our main language is English, so we use tr("Some english text") all over the source code.
We also plan to translate it to several different languages - no problem with that.
Our customer wants to get all phrases from the source code and proofread them.
Of course, we should put those phrases back after proofreading.
How can we accomplish that in a proper way? Maybe Qt Linguist allow to export/import embedded localizable texts?
I guess the customer can just translate English into English and then we can use that English translation, but it's weird.
I would go with Qt's lupdate utility (could be found in Qt's bin directory) that will extract all string literals from your sources into a xml (ts) file. The file can be opened with Linguist tool.
Note, that the utility considers only strings surrounded with tr() macro. Here is luptdate description:
lupdate is part of Qt's Linguist tool chain. It extracts translatable
messages from Qt UI files, C++, Java and JavaScript/QtScript source
code. Extracted messages are stored in textual translation source
files (typically Qt TS XML). New and modified messages can be merged
into existing TS files.
UPDATE:
Another alternative is keeping all string literals definitions in a separate source file and update it once customer has corrected all strings. I believe this happens not so frequently, or even only once, so it would not be worth of much effort with translations etc.
Finally, it looks like I will have to update phrases (embedded into source code) by hand. Actually, it shouldn't take too much time. If I have time to write a script on Python I will update this post.
UPDATE
So, I made everything "by hand" with a little help from Sublime Text 3.
Find all matches in repository folder using the following regular expression (.*)(tr\((\"(.+?)\")\))(.*)
Copy the search results into new document
Using the same regular expression do the search again and replace each match with \4 - this capture group represents text in tr("").
After receiving phrases from the customer after proofreading, it took 3-5 minutes to find differences with diff tool and update phrases in code.
Not a true-programmer way of resolving problems but worked for me and worked pretty fast!
For educational purposes, I would like to build an IDE for PHP coding.
I made a form app and added OpenFileDialog ..(my c# knowledge was useful, because it was easy ... even though without intelisense!)
Loading a file and reading lines from it is basically the same in every language (even PERL).
But my goal is to write homemade intelisense. I don't need info on the richtextBox and the events it generates, endline, EOF, etc, etc.
The problem I have is, how do I handle the data? line for line?
a struct for each line of text file?
looping all the structs in a linked list? ...
while updating the richtextBox?
searching for opening and closing brackets, variables, etc, etc
I think Microsoft stores a SQL type of database in the app project folders.
But how would you keep track of the variables and simulate them in some sort of form?
I would like to know how to handle this efficiently on dynamic text.
Having never thought this through before, it sounds like an interesting challenge.
Personally, I think you'll have to implement a lexical scanner, tokenizing the entire source file into a source tree, with each token also having information about it mapping the token to a line/character inside of the source file.
From there you can see how far you want to go with it - when someone hovers over a token, it can use the context of the code around it to be more intelligent about the "intellisense" you are providing.
Hovering over something would map back to your source tree, which (as you are building it) you would load up with any information that you want to display.
Maybe it's overkill, but it sounds like a fun project.
This sounds to be related to this question:
https://softwareengineering.stackexchange.com/questions/189471/how-do-ide-s-provide-auto-completion-instant-error-checking-and-debugging
The accepted answer of that question recommends this link which I found very interesting:
http://msdn.microsoft.com/en-us/magazine/cc163781.aspx
In a nutshell, most IDEs generate the parse tree from the code and that is what they stores and manage.
I have a book project which I'd like to start sooner than later. This would follow an agile-like publishing workflow, i.e: publish early and often. It is meant to be self-publsihed by me and I'm not really looking to paper-publish it, even though we never know.
If I weren't a geek, I'd probably have already started writting in Word or any other WYSIWYG tool and just export to PDF. However, we know it is not the best solution, and emacs rules my text-editing life, so, the output format should be as simple as possible and be text-based.
I've thought about the following options:
Just use orgmode and export to PDF (orgmode has this feature natively)
Use markdown mode and export to PDF (markdown->LaTeX->PDF should not be hard to setup);
Use something similar to what the guys # Pragmatic Progammers do: A XML + XSLT + LaTeX.
More complex, but much more control over the style.
EDIT: Someone just told me that he uses a combo of Textile+Adobe In Design and the XTags plugin. Not sure how they are glued together though, gotta do some research.
Any other ideas / references ?
I want to start writting as soon as possible. In fact, I already have a draft in an org-formatted file. However, I do want to have and use the full power of LaTex later on to format it the way I want and make it look fabulous :)
Thanks in advance,
Marcelo.
I have done a TON of research on this lately, since I'm planning on starting my own small press soon.
It really depends on what you want your final output to be (PDF, HTML, other?), and what the book is about.
Org mode is great, as I'm sure you know, because it expands as you do. I often write my outlines in org mode, then just fill in the body text when I'm really ready to start writing.
IF it's prose, and you just need some simple divisions (chapters and sections and not much else), org mode -> latex should do you just fine. Then you also have the possibility of org mode -> html
IF you need math in it, you can just write the math right in the org mode file.
If it's really really technical information, docbook might be nice (emacs + nxml), then dockbook 4.5 -> jade -> jadetex -> pdf.
I'd stay away from docbook 5, because it uses FOP to generate PDFs, and the typesetting is really inferior to latex.
BOTTOM LINE: If you want a PDF, use org -> latex, the path of least resistance ;) -- whatever you do, concentrate on the content of the book first, and worry about what it looks like til after.
And why not paper publish? Have you looked at lulu.com? I recently formatted a book with latex, uploaded the pdf to lulu, and had them print it. The quality is pretty good, and definitely worth a look. I have a ton of bookmarks at home about publishing in general, if you're interested.
Typography is hard.
TeX/LaTeX are tools that can get you the best possible results, however they require knowledge about typography to be used correctly--especially with a big document like a book. And I haven't seen any other cheap (=not for professional use) software that would do things correctly automatically. (I haven't seen any professional software, so it is possible they don't do that either)
However, assuming that you'll write your book in some machine-readable format, putting it into TeX/LaTeX should not be very hard: once I had a set of documents in a custom XML format. Proper usage of XSLT, TeXML and LaTeX gave me something I could tweak manually (and this tweaking was necessary!) and get the best possible result.
My advice: prepare content in something that is easy to parse and easy to write in. I'd dismiss XML. Markdown seems to be good choice. This will also allow you to quickly show your work. Then if you decide to make the result better, write some simple script to translate that to TeX (it is not that hard to get basic functionality) and fix things by hand. This might actually be a good exercise to learn TeX.
Don't try to get everything right from the beginning. Firstly get the content, then play with formatting.
If you are really wanting to do online only, I would suggest you use org mode and just stay in HTML. Then you can use CSS to style it however you would like.
That being said, if you really want to output to PDF for technical stuff, I would strongly suggest using Docbook (www.docbook.org). It's made for that, it works great with Emacs.
You have already answered yourself. Not mentioning that you already started writing in org-mode. Org-mode is really extremely powerful and will enable you to publish to PDF and HTML eventually with no effort.
In case of PDF you can take advantage of LaTeX and how org-mode is working with exports. You can include any LaTeX code to your org file. Also IMHO it's way better to write the book/article in org-mode since something becomes even easier than in plain .tex files take for example tables.
Regarding Publishing it's a same story with one single function you can trigger exporting to HTML/PDF and uploading to your server. And notice that you are still using just plain text file which is human readable and very clean.
Org-mode really follows the Emacs philosphy just start using it and it will grow with you.
If you are writing a book, it would certainly be worth the overhead of learning tex.
Even something like,
\documentclass[a4paper,10pt]{book}
\title{SERPA'S BOOK}
\author{SERPA}
\date{\today}
\begin{document}
\maketitle
\tableofcontents
\include{chapterA}
\include{chapterB}
\include{chapterC}
\end{document}
Then, in the same directory have files chapterA.tex, chapterB.tex, chapterC.tex that look like
\chapter{My chapter title}
Lorem ipsum dolor sit amet, consectetur adipiscing elit....
That alone will produce an extremely nice looking document. You can edit each chapter separately and then just compile the main tex file. I think if you try to learn intermediate tools that try to abstract away from tex, you'll only make it more difficult later to do what you actually want, because you will be both fighting tex and an abstraction of tex at the same time.
Best of luck on such an undertaking.
Also, no matter what you do, make sure to use some kind of version control system, such as SVN, to manage your files. It will be worth it.
I would write it in Latex and have an online repository that does nightly compiles to PDF of the 'publish-ready' branch, available to readers.
I would not start with using LaTeX these days. TeX input is unstructured and the only thing you can get out of TeX input is PDF. If you need HTML or anything else, you are screwed.
Use something structured, such as XML (DocBook is a good suggestion) or define your own XML subset as you need it. Use XSLT to transform it into something usable (HTML etc.) That way you are set for the future.
Depending on your typographical needs, you can then use TeX as a backend processor, or XSLT or whatever.
Also, have a look at ConTeXt, it can read XML directly and has great typography!