more font optimization: subset by style

[alixander]

currently we subset fonts by collecting all characters and encoding, in every style, those characters.

we can further optimize by subsetting each character WITH its style, so that the font encoding for bold only includes the bolded characters, the font encoding for italic only includes the italics, etc.

related initial optimization: https://github.com/terrastruct/d2/pull/1089

[alixander]

@donglixiaoche can you leave a comment here so i can assign you? (github won't let me assign otherwise)

[donglixiaoche]

@alixander coooopy, wanna give this issue a shot

[donglixiaoche]

@alixander when will a text be semibold? there's no such a flag inside Text struct indicates it's semibold, only bold flag

type Text struct {
    Label      string `json:"label"`
    FontSize   int    `json:"fontSize"`
    FontFamily string `json:"fontFamily"`
    Language   string `json:"language"`
    Color      string `json:"color"`

    Italic    bool `json:"italic"`
    Bold      bool `json:"bold"`
    Underline bool `json:"underline"`

    LabelWidth  int    `json:"labelWidth"`
    LabelHeight int    `json:"labelHeight"`
    LabelFill   string `json:"labelFill,omitempty"`
}

[donglixiaoche]

ohh i see, when there is markdown inside diagram, right? when will there be a markdown ? an explanation?

[donglixiaoche]

https://github.com/terrastruct/d2/blob/feae388f19b72bb9c90de7452472bfd154bcd86c/d2renderers/d2svg/d2svg.go#L1373-L1394

and why do we need regular font for md class? below these lines, we also add semibold font for md class