TRA file format

[mickabouille]

Hi, I was writing a small tool/script to work on weidu tra files and expected to be, well, not that hard.

But then after naively pumping out code I noticed I had forgotten things.

This is not enough

```

@<id> = ~text~{~text~}

```

(with {...} meaning optional)

Even when allowing both ~ and " as string delimiter (those are the only two, promise?) and comments.

Because I had TRAs with `[sound]` too!

```

@<id> = ~text~{[sound]} {~text~}  {[sound]}

```

And then I found some TRA files with **negative** indexes

Well, looking at the weidu doc, I couldn't find a description of the format, not in the sens that we have a semi-formal description of the .D format. Well it seems maybe it was an extension of .D files themselves(?)

Or do we have one and I couldn't find it? Do I have everything or am I still missing parts:

- comments with `//

- Do we have `/* */` too? Can we put those between `@123` and `=` or somewhere else in the middle of an entry?

- an id in the form `@<number>` including negative numbers

- strings enclosed in tildas (with no tildas inside) or double quotes (with no double quotes inside)

- one or two string with zero, one or two sound labels

- Uh what characters are accepted in the sound label by the way? Probably not `]` I suspect, but any other restriction?

This was supposed to be a peaceful, relaxing, leisurely activity (and then "let's write a LALR grammar"...)

Edited August 2 by mickabouille

[argent77]

1 hour ago, mickabouille said:

- strings enclosed in tildas (with no tildas inside) or double quotes (with no double quotes inside)

That's not all. Strings can also be enclosed in percent signs (%) or five consecutive tildes (~~~~~). And string concatenation is most likely also supported (string1 ^ string2).

The translation string syntax is mostly covered in the WeiDU D format description (see the Text and String subsections).

[medic2000]

Is there any TRA file editor that allows us only deal with text and not mess any of the special characters?

[lynx]

Yes, for mods that use burner1024's converter to PO files — you can use any proper translation tool then, including his instance of (pootle). A good way to benefit from a translation memory.

[mickabouille]

11 hours ago, argent77 said:

That's not all. Strings can also be enclosed in percent signs (%) or five consecutive tildes (~~~~~). And string concatenation is most likely also supported (string1 ^ string2).

The translation string syntax is mostly covered in the WeiDU D format description (see the Text and String subsections).

I was hoping this unnecessary cruft would have been banned from TRA to make it an inert data format

Wait does that mean you can do that too?

@2 = @1 ^ @109 ^ #234

@1 = @2 ^ @109

 

EDIT: and, eh... This five tildes things is a pain, talk about grammar conflicts

Edited August 3 by mickabouille

[argent77]

1 hour ago, mickabouille said:

Wait does that mean you can do that too?

@2 = @1 ^ @109 ^ #234

@1 = @2 ^ @109

No, you can only concatenate strings. E.g.

@1 = "This is" ^ ~ a string.~
@2 = %This is% ^ ~~~~~ another string.~~~~~

[mickabouille]

Well, after testing, I'm somewhat relieved to confirm that you can't do

@10 = ~something~ ^ @11

BUT you can do

@10 = @11

@10 = #1234 //also works

And

 

@10 = @11

@11 = @11

is just a stack overflow

EDIT: mostly in sync

Edited August 3 by mickabouille

[mickabouille]

> No, you can only concatenate strings.

Ah ok, this is what I call string literals.

Edited August 3 by mickabouille

[mickabouille]

And of course you CAN put multiple entries on the same line (I guess you could put your entire TRA file on a single line).

And you can't do

@12 = ~some text~ @13

which I admit would be confusing semantically.

And, (wooot) this works!

@11  = ~Something~ /* comment */ ~Something else~

(sob )

Edited August 3 by mickabouille

[medic2000]

5 hours ago, lynx said:

Yes, for mods that use burner1024's converter to PO files — you can use any proper translation tool then, including his instance of (pootle). A good way to benefit from a translation memory.

I didn't understand any of this.

- Who is burner1024?

- What is PO file? Why would i convert to PO file?

- What is pootle?

- How will I know if a mod uses burner1024's converter to PO files? BG1 NPC Project?

- Any proper translation tool? For example? I only know DeepL.

Edited August 3 by medic2000

[mickabouille]

So, I just tested

`@11 = ~~~~~~~~~~~~~~`

(How far am I gone?)

and weidu doesn't want anything to do with it (it should mean `~~~~`)

Even `@11 = ~~~~~~~~~~~` doesn't work (it should mean `~`

On the other hand `@11 = ~~~~~ ~~~~ ~~~~~` works and shows `~~~~`

So now my parser is more correct than weidu... which is a problem because I don't intend to RE the exact behavior.

(Seems I am neither on Earth nor on Faerûn, that's maybe be the Fugue plane?)

Edited August 3 by mickabouille

[kjeron]

33 minutes ago, mickabouille said:

`@11 = ~~~~~~~~~~~~~~`

and weidu doesn't want anything to do with it (it should mean `~~~~`)

Even `@11 = ~~~~~~~~~~~` doesn't work (it should mean `~`

On the other hand `@11 = ~~~~~ ~~~~ ~~~~~` works and shows `~~~~`

So now my parser is more correct than weidu... which is a problem because I don't intend to RE the exact behavior.

You have an empty string "~~~~~~~~~~" followed by four tildes, not four tildes within two instances of "~~~~~".  A string ends at the first instance of it's delimiter.   "~~~~~~~~~~~~" (12x) will work, because it reads two empty strings: "~~~~~~~~~~" and "~~".  If you want tildes within a string like that, you need to use either " or % as the delimiter, or use a variable.

[mickabouille]

Doesn't the weidu doc say "five consecutive tildes on each side"?

Edited August 3 by mickabouille

[argent77]

@1 = ~~~~~~~~~~~~ // (12x)
seems to work because WeiDU interprets it as two consecutive string literals (male and female string) without having any whitespace in between.

It is functionally identical to
@1 = ~~~~~~~~~~""
or
@1 = ""~~~~~~~~~~
or even
@1 = ~~~~~~~~~~[]~~[]  // two strings with empty sound references

[mickabouille]

That's disturbing.

Anyway, this five tildes things has no real use (given we already bear the cost of concatenation support) and it I only (barely) care about it because one of my test case (the tra-ified bg1ee tlk) contains

@30    = ~~~~~abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ1234567890!@#$%^&*() -_=+\|]}[{;:'",<.>/?–— ÀÁÂÃÄÅÆÇÈÉÊËÌÍÎÏÐÑÒÓÔÕÖÙÚÛÜÝÞßàáâãäåæçèéêëìíîïðñòóôõöøùúûüýþÿœŒ€£~°µ~~~~~

 

So I'll probably disregard the more complex cases.

Edited August 4 by mickabouille