Loading fonts from memory #199

klausweiss · 2022-08-04T17:37:02Z

I was looking into loading fonts from memory in addition to loading from file, in order to be able to easily distribute monomer programs as standalone binaries. Since it isn't possible at the time of writing, would you consider adding this to monomer? I attempted implementing it.

What I have in this PR doesn't work unfortunately. It compiles, but the text doesn't show up (see the labels in the modified todo example below). It's my first time using c2hs, so I must have got something wrong. I'm out of expertise, so I'm calling for help if anyone is interested in doing so.

The todo changes (and an additional dependency) are only there for testing easily. These changes should be reverted when it's finally working.

Screenshot

klausweiss · 2022-08-04T17:37:31Z

src/Monomer/Main/Types.hs

+{- |
+Alias for 'appFontDefFile' for backwards compatibility.
+-}
+appFontDef = appFontDefFile
+{-# DEPRECATED appFontDef "Use appFontDefFile directly" #-}


If the changes here get through, I propose deprecating appFontDef in the light of having 2 distinct ways of loading fonts.

If we go down this route, we'll also need to update the examples and docs code to use appFontDefFile instead of appFontDef.

@klausweiss I'd prefer to keep the existing appFontDef name as is, and add the new memory based one as appFontDefMem. The main reason is to avoid breaking existing code. A secondary reason is this would make it work similarly to the image widget, that has image for path based images and imageMem for memory based ones.

If you prefer to keep it as appFontDefMemory, it works great for me too.

Ok, sounds good (re deprecation).

Renamed appFontDefMemory back to appFontDefMem.

klausweiss · 2022-08-04T21:00:17Z

src/Monomer/Graphics/FontManager.hs

  if res >= 0
-    then return $ path : fonts
+    then return $ name : fonts


Since the values inside the resulting IO [Text] weren't used and the fonts loaded from memory don't have a path identifying them, this was changed to return a list of font names instead.

Makes sense! 👍

fjvallarino · 2022-08-05T03:42:08Z

@klausweiss I'll take a look during the weekend.

What I would check first is if the loading functions are returning a valid status code (i.e., they are not failing to load the embedded binary content). I'd also validate that the file embedding is actually happening. Finally, I'd check the chs bindings (I added a comment there).

fjvallarino · 2022-08-05T03:42:44Z

src/Monomer/Graphics/FFI.chs

@@ -118,6 +119,23 @@ withText t = useAsCString (T.encodeUtf8 t)
 withNull :: (Ptr a -> b) -> b
 withNull f = f nullPtr

+allocCUString :: ByteString -> ((Ptr CUChar, CInt) -> IO a) -> IO a


@klausweiss I'm thinking maybe something like this could work (I have not tested it, and it may not compile as it stands):

withByteString :: ByteString -> (CString -> IO b) -> IO b withByteString bs = useAsCString bs {# fun unsafe fmCreateFontMem {`FMContext', withCString*`Text', withByteString*`ByteString'&} -> `Int' #}

Unfortunately the C API requires it to use Ptr CUChar instead of Ptr CChar, which CString uses. I've made some progress (see this comment and the code I pushed), but that's still not it.

klausweiss · 2022-08-05T15:01:58Z

Thanks @fjvallarino!

What I would check first is if the loading functions are returning a valid status code (i.e., they are not failing to load the embedded binary content).

They do. 0, 1, 2, 3, ... for each consecutive call to the C function.

I'd also validate that the file embedding is actually happening.

Confirmed it's working, exactly the same bytes as in the input file.

Finally, I'd check the chs bindings (I added a comment there).

I've made some progress here. I've tried to use what's in nanovg-hs for this bit: https://github.com/cocreature/nanovg-hs/blob/cc8dfa0dc18a0792786c973b4e9a232fa7d3ecfd/src/NanoVG/Internal/FFIHelpers.hs#L24-L26

Now it randomly renders some dots where the text should be on labels:

Screenshot

I tried explicitly allocating memory for the bytestring (updated the PR with this code), but the results are same as above - randomly rendered dots.

klausweiss · 2022-08-05T16:28:23Z

Ok, I got this. I think it's the nanovg-hs code that might be a bit off.
I modified nanovg-hs to use the same trick with allocating memory for the font on my own to prevent it from being garbage-collected:

copyCUStringLenMemory (from, len) =
  let intLen = fromIntegral len
  in do
    to <- mallocBytes intLen
    copyBytes to from intLen
    return (to, len)

and made the NVGcontext free the font upon destruction. The results are shown in the image below (:

Screenshot

I'll open a PR to nanovg-hs to fix this later.

This reverts commit 06d061a.

klausweiss

Now that the nanovg fix has been released, I'm happy to open this PR for review.

I left some comments in places I'm unsure about.

klausweiss · 2022-08-21T11:03:34Z

src/Monomer/Graphics/Types.hs

+data FontDef
+  = FontDefFile
+    { _fntFontName :: !Text  -- ^ The logic name. Will be used when defining styles.
+    , _fntFontPath :: !Text  -- ^ The path in the filesystem.
+    }
+  | FontDefMem
+    { _fntFontName :: !Text         -- ^ The logic name. Will be used when defining styles.
+    , _fntFontBytes :: !ByteString  -- ^ The bytes of the loaded font.
+    }
+  deriving (Eq, Show, Generic)


I'd love to use this

data FontDef = FontDefFile { _fntName :: !Text -- ^ The logic name. Will be used when defining styles. , _fntPath :: !Text -- ^ The path in the filesystem. } | FontDefMem { _fntName :: !Text -- ^ The logic name. Will be used when defining styles. , _fntBytes :: !ByteString -- ^ The bytes of the loaded font. } deriving (Eq, Show, Generic)

but for reasons I don't understand (some lens stuff) I can't:

[ 21 of 104] Compiling Monomer.Graphics.Lens [Monomer.Graphics.Types changed] .../monomer/src/Monomer/Graphics/Lens.hs:24:1: error: • Could not deduce (Applicative f) arising from a use of ‘pure’ from the context: Functor f bound by the type signature for: path :: Control.Lens.Type.Lens' FontDef Data.Text.Internal.Text at src/Monomer/Graphics/Lens.hs:24:1-42 Possible fix: add (Applicative f) to the context of the type signature for: path :: Control.Lens.Type.Lens' FontDef Data.Text.Internal.Text • In the expression: pure ((FontDefMem x1_acDU) x2_acDV) In an equation for ‘path’: path _ (FontDefMem x1_acDU x2_acDV) = pure ((FontDefMem x1_acDU) x2_acDV) In the instance declaration for ‘HasPath FontDef Data.Text.Internal.Text’ | 24 | makeLensesWith abbreviatedFields ''FontDef | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Also note this change creates traversals, not lenses, which is something to watch out for, see GHCI:

λ: FontDefMem "name" "Bytes" ^. fontPath ""

I'm open to alternatives, but given how small these constructors are, I'm fine with leaving as is.

I'm fine with the naming you chose. I guess the error is caused by a previous Lens/Prism generated in https://github.com/fjvallarino/monomer/blob/main/src/Monomer/Lens.hs.

Alternatively, since this constructors are only used in a couple of places, you can use:

data FontDef = FontDefFile !Text !Text -- ^ Associates a font name with a filesystem path. | FontDefMem !Text !ByteString -- ^ Associates a font name with a font loaded in memory. deriving (Eq, Show, Generic)

Then, in NanoVGRenderer, you need to deconstruct:

... createFont (FontDefFile name path) = VG.createFont c name (VG.FileName path) createFont (FontDefMem name bytes) = VG.createFontMem c name bytes ...

Thanks, I went with what you suggested. name was needed outside of the createFont function in both NanoVGRenderer and FontManager, so I hand-crafted a HasName instance for FontDef, since it can't be generated anymore due to no record syntax.

klausweiss · 2022-08-21T11:06:44Z

src/Monomer/Main/Types.hs

+{- |
+Alias for 'appFontDefFile' for backwards compatibility.
+-}
+appFontDef = appFontDefFile
+{-# DEPRECATED appFontDef "Use appFontDefFile directly" #-}


If we go down this route, we'll also need to update the examples and docs code to use appFontDefFile instead of appFontDef.

Dretch · 2022-09-10T16:54:17Z

I had a play around locally and it works for me 😄

I wonder if there should be a pointer in the docs to https://hackage.haskell.org/package/file-embed-0.0.15.0/docs/Data-FileEmbed.html or similar (this is what I used and I assume this is intended).

Also there is appWindowIcon that might benefit from something similar, to enable fully self-contained binaries.

fjvallarino · 2022-09-10T20:55:58Z

@klausweiss I'm really sorry I missed this. I didn't see you had switched it to a PR ready for review.

I added a couple of comments. The PR looks great!

fjvallarino · 2022-09-10T21:06:43Z

src/Monomer/Main/Types.hs

+
+{-|
+Available fonts to the application, loaded from the bytes in memory. 
+Specifying no fonts will make it impossible to render text.


Maybe adding a short example about how to embed a font would be a good idea (the text I added is just a sample, change it as you see fit).

{-| Available fonts to the application, loaded from the bytes in memory. Specifying no fonts will make it impossible to render text. One use case for this function is to embed fonts in the application, without the need to distribute the font files. The [file-embed](https://hackage.haskell.org/package/file-embed-0.0.15.0/docs/Data-FileEmbed.html) library can be used for this. @ appFontDefMemory "memoryFont" $(embedFile "dirName/fileName") @ -} appFontDefMemory :: Text -> ByteString -> AppConfig e

This reverts commit 5403752.

klausweiss

@klausweiss I'm really sorry I missed this. I didn't see you had switched it to a PR ready for review.

@fjvallarino no worries, I'll make sure to mention you next time I do this (:

Also there is appWindowIcon that might benefit from something similar, to enable fully self-contained binaries.

@Dretch sounds good to me, happy to do this in a separate PR. Will need a couple days to find time for that though.

klausweiss · 2022-09-11T07:08:56Z

src/Monomer/Main/Types.hs

+{- |
+Alias for 'appFontDefFile' for backwards compatibility.
+-}
+appFontDef = appFontDefFile
+{-# DEPRECATED appFontDef "Use appFontDefFile directly" #-}


Ok, sounds good (re deprecation).

Renamed appFontDefMemory back to appFontDefMem.

klausweiss · 2022-09-11T07:12:33Z

src/Monomer/Graphics/Types.hs

+data FontDef
+  = FontDefFile
+    { _fntFontName :: !Text  -- ^ The logic name. Will be used when defining styles.
+    , _fntFontPath :: !Text  -- ^ The path in the filesystem.
+    }
+  | FontDefMem
+    { _fntFontName :: !Text         -- ^ The logic name. Will be used when defining styles.
+    , _fntFontBytes :: !ByteString  -- ^ The bytes of the loaded font.
+    }
+  deriving (Eq, Show, Generic)


Thanks, I went with what you suggested. name was needed outside of the createFont function in both NanoVGRenderer and FontManager, so I hand-crafted a HasName instance for FontDef, since it can't be generated anymore due to no record syntax.

klausweiss · 2022-09-11T07:13:46Z

nix/monomer.nix

-      rev = "cc8dfa0dc18a0792786c973b4e9a232fa7d3ecfd";
-      sha256 = "0vvj4l2dfjqspl80bwq4vkcql5p7s5a7l1cv7vajkak0vn1ryy70";
+      rev = "283516a337c4d3606555728df0a39294e78a7cdf";
+      sha256 = "1vggli76ijqmx633yix4yg5dv58a14p8561jnprjc061sjngphzv";


For posterity: computed with

nix-prefetch-url --unpack https://github.com/cocreature/nanovg-hs/archive/283516a337c4d3606555728df0a39294e78a7cdf.tar.gz

Nice! I had not yet needed to update Nix dependencies; it's good to know.

fjvallarino · 2022-09-11T08:08:38Z

src/Monomer/Graphics/FontManager.hs

-    name = fontDef ^. fontName
-    createFont FontDefFile{} = fmCreateFont ctx name (fontDef ^. fontPath)
-    createFont FontDefMem{} = fmCreateFontMem ctx name (fontDef ^. fontBytes)
+    name = fontDef ^. L.name


Looks good!

Just to be clear: if you prefer your previous version with the records, go ahead with it; it works well for me.

fjvallarino · 2022-09-11T08:14:16Z

@klausweiss you did mention switching from Draft to Ready for Review in a previous comment, I just missed it :)

All looks good for me. I added a comment about choosing between your original design for the FontDef type or the alternative I proposed and, based on what you decide, I'm ready for merging.

Thanks!

klausweiss · 2022-09-11T09:50:13Z

All looks good for me. I added a comment about choosing between your original design for the FontDef type or the alternative I proposed and, based on what you decide, I'm ready for merging.

Ok. I'll revert the change for the sake of not having to hand-craft the lens instance and am happy to merge afterwards. Thanks for the assistance!

This reverts commit 03d16ce.

fjvallarino · 2022-09-11T10:28:23Z

Thank you for the patience!

klausweiss · 2022-12-29T21:18:04Z

Also there is appWindowIcon that might benefit from something similar, to enable fully self-contained binaries.

@Dretch sounds good to me, happy to do this in a separate PR. Will need a couple days to find time for that though.

Sorry @Dretch, I didn't get round to doing this eventually. It's free for anyone to pick up.

klausweiss added 5 commits August 3, 2022 17:27

Compiles, doesn't work

32e067a

s/FontDefPath/FontDefFile/g

fac20ed

Some ffi changes

13f55dd

Rename toUString

5cbbe4b

Demo in the todo app. TO BE REVERTED

06d061a

klausweiss commented Aug 4, 2022

View reviewed changes

Alloc an unsigned char* array for font

ec4bf11

klausweiss commented Aug 4, 2022

View reviewed changes

klausweiss added 2 commits August 4, 2022 23:15

Remove unused language extensions

e439230

Remove unused and sort imports

1e4bd8b

fjvallarino reviewed Aug 5, 2022

View reviewed changes

Use useAsCStringLen and malloc manually

d6f11a0

Use nanovg from klausweiss/nanovg-hs (TO BE REVERTED)

cf7839c

klausweiss mentioned this pull request Aug 7, 2022

Fix createFontMem to not use garbage-collected memory cocreature/nanovg-hs#17

Merged

klausweiss added 7 commits August 7, 2022 21:46

Use newer klausweiss/nanovg-hs (TO BE REVERTED)

9e3bac9

Use nanovg 0.8.1 from hackage

0ee513d

Revert "Demo in the todo app. TO BE REVERTED"

85ce069

This reverts commit 06d061a.

Add documentation to some functions in FFI.hs

cf68aea

Unused import

ced04e2

Improve appFontDef docs

7112d1e

s/appFontDefMem/appFontDefMemory

5403752

klausweiss commented Aug 21, 2022

View reviewed changes

klausweiss marked this pull request as ready for review August 21, 2022 12:58

fjvallarino reviewed Sep 10, 2022

View reviewed changes

Bump nanovg-hs to 0.8.1 in monomer.nix

7f1db12

klausweiss added 4 commits September 11, 2022 07:42

Revert "s/appFontDefMem/appFontDefMemory"

5cd391b

This reverts commit 5403752.

Reference file-embed in appFontDefMem docs

375c5cb

s/appFontDefFile/appFontDef

7e15883

Stop using record syntax for FontDef

03d16ce

klausweiss commented Sep 11, 2022

View reviewed changes

fjvallarino reviewed Sep 11, 2022

View reviewed changes

Revert "Stop using record syntax for FontDef"

7880379

This reverts commit 03d16ce.

fjvallarino merged commit 619ac2d into fjvallarino:main Sep 11, 2022

klausweiss deleted the feature/appFontDefMem branch September 11, 2022 10:31

fjvallarino mentioned this pull request May 23, 2023

PR proposal: run without config / files (package at least roboto into the lib), combinators above file embed, UI-only run #288

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Loading fonts from memory #199

Loading fonts from memory #199

klausweiss commented Aug 4, 2022 •

edited

Loading

klausweiss Aug 4, 2022 •

edited

Loading

klausweiss Aug 21, 2022

fjvallarino Sep 10, 2022

klausweiss Sep 11, 2022 •

edited

Loading

klausweiss Aug 4, 2022

fjvallarino Sep 10, 2022

fjvallarino commented Aug 5, 2022

fjvallarino Aug 5, 2022

klausweiss Aug 5, 2022 •

edited

Loading

klausweiss commented Aug 5, 2022 •

edited

Loading

klausweiss commented Aug 5, 2022 •

edited

Loading

klausweiss left a comment

klausweiss Aug 21, 2022

klausweiss Aug 21, 2022

fjvallarino Sep 10, 2022 •

edited

Loading

klausweiss Sep 11, 2022

klausweiss Aug 21, 2022

Dretch commented Sep 10, 2022

fjvallarino commented Sep 10, 2022

fjvallarino Sep 10, 2022

klausweiss left a comment

klausweiss Sep 11, 2022 •

edited

Loading

klausweiss Sep 11, 2022

klausweiss Sep 11, 2022

fjvallarino Sep 11, 2022

fjvallarino Sep 11, 2022

fjvallarino commented Sep 11, 2022

klausweiss commented Sep 11, 2022

fjvallarino commented Sep 11, 2022

klausweiss commented Dec 29, 2022

Loading fonts from memory #199

Loading fonts from memory #199

Conversation

klausweiss commented Aug 4, 2022 • edited Loading

klausweiss Aug 4, 2022 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

klausweiss Sep 11, 2022 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

fjvallarino commented Aug 5, 2022

Choose a reason for hiding this comment

klausweiss Aug 5, 2022 • edited Loading

Choose a reason for hiding this comment

klausweiss commented Aug 5, 2022 • edited Loading

klausweiss commented Aug 5, 2022 • edited Loading

klausweiss left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

fjvallarino Sep 10, 2022 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Dretch commented Sep 10, 2022

fjvallarino commented Sep 10, 2022

Choose a reason for hiding this comment

klausweiss left a comment

Choose a reason for hiding this comment

klausweiss Sep 11, 2022 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

fjvallarino commented Sep 11, 2022

klausweiss commented Sep 11, 2022

fjvallarino commented Sep 11, 2022

klausweiss commented Dec 29, 2022

klausweiss commented Aug 4, 2022 •

edited

Loading

klausweiss Aug 4, 2022 •

edited

Loading

klausweiss Sep 11, 2022 •

edited

Loading

klausweiss Aug 5, 2022 •

edited

Loading

klausweiss commented Aug 5, 2022 •

edited

Loading

klausweiss commented Aug 5, 2022 •

edited

Loading

fjvallarino Sep 10, 2022 •

edited

Loading

klausweiss Sep 11, 2022 •

edited

Loading