The Taamey D font for Biblical Hebrew

Author: Ben Denckla

Font version: 0.921

Document version: 2023-03-13 / כ׳ בַּאֲדָר תשפ״ג

Introduction

Taamey D is derived from Taamey Frank Clm, a font designed by Yoram Gnat, of blessed memory. Taamey D’s goal is to support every word in the following two Hebrew Bible datasets:

• Uxlc (Unicode xml Leningrad Codex), whose canonical edition is hosted at tanach.us
• Mam (Miqra according to the Masorah), whose canonical edition is hosted at Hebrew Wikisource

This document has the following sections:

• Distinctive features
• Challenging cases
• Floating ḥolam ḥaser
• Notes on the limited corpus

Many people have helped me develop this font and develop my Hebrew enough to develop this font. Thanks to:

• Simon Cozens, for “collidoscope” and his other useful software for font development, and for helping me get started using it.
• Scott-Martin Kosofsky, for answering my questions about Hebrew fonts & typesetting.
• John Hudson, for answering my questions about Hebrew fonts & typesetting. I’d also like to thank John for his Sbl Hebrew Font User Manual, the best reference for how to use Unicode to tell a font what you want in Biblical Hebrew.
• Avi Kadish, for teaching me much of what I know about the Hebrew Bible and for giving me the opportunity to work on mam.
• Chris Kimball, for giving me the opportunity to work on uxlc.
• Ron Tabeka, for teaching me Hebrew & chanting.

Distinctive features

This section describes the most distinctive features of Taamey D, as compared to Taamey Frank.

Upside-down qamats qatan

As in many fonts, in Taamey D it can be hard to distinguish qamats qatan from normal qamats. (Despite what its name might suggest, qamats qatan is slightly taller.) Taamey D provides a stylistic set, ss03, in which the qamats qatan shape is upside-down. This makes it easy to distinguish. Here is a word [1] rendered both ways:

(This word is from mam since uxlc does not use qamats qatan.) This upside-down feature is mainly intended for use within this document, but it might find use in other places such as teaching materials.

The ḥataf qamats qatan vowel

Taamey D provides a stylistic set, ss01, in which the glyph for ḥataf qamats matches qamats qatan. Below I render an example word [2] in default Taamey D, and in Taamey D with ss01 selected. Corresponding upside-down versions are provided as well. This word’s third qamats is a normal qamats, which helps as a reference. Letters with matching qamats shapes are highlighted.

The qamats qatan vowel with a short meteg

mam records two “qamats variants” of some words. These are words having a qamats that is:

• a qamats qatan, according to theory (abbrev. ד (dalet))
• a normal qamats, according to Sephardic practice (abbrev. ס (samekh))

For example, below are mam’s two variant pointings of a word [3]. The pointings are rendered in both default and upside-down Taamey D. The mem whose qamats varies is highlighted. Below I am just showing the “qamats variants” feature of mam: I am not showing the short meteg feature of Taamey D yet.

Taamey D provides a stylistic set, ss02, in which the glyph for qamats qatan has a short meteg. This allows both variants to be presented on a single word. A qamats qatan with a small meteg represents a qamats that is qatan only in the theoretical variant. Below is our example theoretical variant shown with ss02 selected (in addition to ss01, for ḥataf qamats qatan):

Distinct atnaḥ hafukh & yeraḥ ben yomo

This phrase [4] shows that Taamey D distinguishes yeraḥ ben yomo from atnaḥ hafukh:

כִּ֤י אֵ֪ין בְּפִ֡יהוּ נְכוֹנָה֮ קִרְבָּ֢ם הַ֫וּ֥וֹת

Taamey Frank & many other fonts do not distinguish these accents. Such fonts give an atnaḥ hafukh shape for both code points. This shortcoming likely stems from the influence of bhs.

Bhs is so influential that even its shortcomings still reverberate through the Biblical world, showing up not only in other editions of the Hebrew Bible but also in fonts! Bhs identifies an accent which it calls “Galgal vel Jèraḥ” (“galgal or yeraḥ [ben yomo]”). So far so good, since yeraḥ ben yomo is also known as galgal. But then things fall short in the following ways:

• Bhs gives an atnaḥ hafukh shape to “Galgal vel Jèraḥ.”
• Bhs does not identify a distinct atnaḥ hafukh accent.

So, many fonts did the following:

• Following bhs, they gave an atnaḥ hafukh shape to yeraḥ ben yomo.
• Quite reasonably, they gave an atnaḥ hafukh shape to atnaḥ hafukh.

Uxlc used to follow wlc, which follows bhs closely. So, uxlc used to use only the yeraḥ ben yomo code point. Uxlc started to distinguish yeraḥ ben yomo from atnaḥ hafukh in version 1.3, released on 19 October 2021. To be able to see this distinction, tanach.us (uxlc’s canonical edition) switched its default font to Taamey D. Formerly it used Sbl Hebrew which, like many fonts, prevents this distinction from being seen, forcing even editions that make this distinction to still look like bhs!

Challenging cases

Below, I will render challenging cases in Taamey D & three other fonts:

• Taamey Frank
• Sbl Hebrew
• Ezra Sil

In these challenging cases, often at least one of those three fonts struggles. Typically that takes the form of a collision or, if not actually a collision, an infelicitous crowding of marks. I am not showing these problems to ridicule these fonts. I have the utmost respect for these fine fonts & their authors. I am showing these problems because one of the main goals of Taamey D was to avoid such collisions & crowding.

It is probably not coincidental that many of the challenging cases are not present in uxlc. I suspect the corpus of concern for these fonts was limited to uxlc or some other Unicode version of wlc. This would make sense because, particularly in the non-Hebrew-speaking world:

• Uxlc not only was, but remains the only well-known Unicode Hebrew Bible.
• Bhs not only was, but remains hugely influential. (Uxlc used to follow wlc, which follows bhs closely.)

The qamats qatan vowel with merkha or munaḥ

Uxlc does not use qamats qatan. In mam, meteg is the only below-mark that qamats qatan appears with regularly. But, in a handful of words in mam, merkha or munaḥ appears with qamats qatan. Probably because these words are so rare, some fonts struggle with them. Below, in Taamey D, are two such words. These words are followed by the relevant pointed letters of those examples, rendered in all 4 of our standard fonts. (I will point out that the gray maqaf below is intentional, but I will not elaborate on it here because it is irrelevant to the point at hand.)

כׇּ֥ל־עַצְמוֹתַ֨י [5] עַל־כׇּל־קׇ֣דָשַׁ֔י [6]

Only Ezra Sil & Taamey D are collision-free. (Actually Ezra Sil does have a collision, between the munaḥ & the qof, but that is unrelated to the qamats being qatan.) When Taamey Frank & Sbl Hebrew misbehave similarly, as they do here, it is not surprising, since the OpenType logic for Taamey Frank is derived from Sbl Hebrew.

The geresh-tg & gershayim-tg compound accents

Five words in mam have, on the same letter, telisha gedolah preceded by either geresh or gershayim. (Only 3 such words exist in uxlc.) Some of these words, particularly those with geresh, prove challenging to some fonts. Here are these words in our standard 4 fonts:

(References are [7], [8], [9], [10], and [11].) Two of the 5 words are not stressed on the first syllable: uleleh & qirvu. For these two words, mam uses a clever but novel stress-helper notation. Taamey Frank & Sbl Hebrew do not support that notation well on uleleh.

Note that the floating ḥolam ḥaser feature (discussed below) lessens the challenge of zot above, because it lessens competition between the dot & the accents. (All 4 fonts have this feature.)

Competition between ḥolam ḥaser & other above-marks

Particularly on narrow letters, ḥolam ḥaser & other above-marks compete for space. Taamey D does its best to move the ḥolam ḥaser out of the way, to the left, in such cases. Many other fonts are not so savvy, and suffer collisions as a result. Note that the examples below all involve stress helpers for accents other than pashta. Stress helpers for accents other than pashta are rare in uxlc. As I mentioned above, I am not surprised that many fonts would struggle with words not present in uxlc, since I suspect that the corpus of concern for many fonts was limited to a Unicode version of wlc. (References are [12], [13], and [14].)

For another example of competition between ḥolam ḥaser & another above-mark, see the section below on stress-helper zarqa.

The revia mugrash accent pair

The revia mugrash accent pair is only present in the sifrei emet system of cantillation. It is composed of revia & the prepositive accent called geresh muqdam. When the first syllable is the stressed syllable, it can be challenging to mediate the competition for space between these two marks if they need to both appear above a narrow letter like gimel. For example [15]:

The lamed ascender

The lamed ascender creates a variety of challenges. In the examples below, Ezra Sil struggles with the ascender. The other fonts do okay, though one might say that Sbl Hebrew’s gershayim comes a little too close for comfort to the ascender. (References are [16], [17], and [18].)

Sometimes the ascender causes a problem indirectly rather than directly. In the word [19] shown below, in Sbl Hebrew & Ezra Sil, the ascender has caused the pashta to be placed far to the left, which may be fine for true (postpositive) pashta but can cause problems for stress-helper pashta. In this case, a stress-helper pashta collides with a ḥolam male dot. Taamey Frank’s stress-helper pashta does not collide with the ḥolam male dot, but it collides directly with the ascender.

Taamey D avoids the problem by placing stress-helper pashta roughly in the center of the lamed, behind the ascender. This also makes ownership of the stress-helper pashta clearer. Note that stress-helper pashta is common in uxlc, unlike other stress helpers, which are rare in uxlc. (Aside: unlike the other three fonts, Taamey Frank does not place furtive pataḥ “early.”)

The deḥi accent

The sifrei emet accent deḥi can be challenging, particularly since it can appear on the same consonant as some other below-marks, most often with meteg but with others such as munaḥ, too. Taamey Frank & Sbl Hebrew seem to be trying to place deḥi up against the vowel rather than leaving it in a prepositive position. This placement is not only a departure from tradition, but also harder to implement! Below are two words with deḥi, shown in Taamey D, followed by their pointed first letters shown in our standard 4 fonts of interest.

לֵֽ֭אלֹהִים [20] שֶׁ֣֭עֹמְדִים [21]

The ayin descender

In the word [22] below, you can see how, in Sbl Hebrew, an ayin with a descender, if followed by an overburdened narrow letter (here zayin), can result in a collision. To avoid the collision, the font might use a larger advance on the ayin in such contexts. Or it might include such contexts in the conditions under which descender-free ayin is used.

Fonts are software, and as with all software, every feature added creates opportunities for new bugs to creep in. In the word [23] below, we can see that Sbl Hebrew neglected to “tell” the descender-free ayin glyph how to place u-zarqa, and as a result it appears as a postpositive on descender-free ayin. Taamey Frank, though it has only one “flavor” of ayin, inherited the vestiges of this feature from Sbl Hebrew, one of the vestiges being this bug.

(The different zarqa shape in Ezra Sil is discussed in the next section. Also, the next section has more information on the confusingly-named u-zarqa code point.)

Stress-helper zarqa

The u-zinor code point is used for true (postpositive) zarqa and the sifrei emet accent tsinor. It is slightly odd that this code point is named after its sifrei emet name. (E.g., Unicode uses u-tipeha not u-tarha.) It is also slightly odd that it is not spelled u-tsinor. (E.g., Unicode uses u-tsadi & u-tsere.) So, the name “u-zinor” is slightly odd in a couple of ways, but no big deal.

The big deal is that there is a code point named u-zarqa that is not used (and should not be used) for true zarqa. Its name is widely agreed to be an error. But due to the nature of Unicode, its name will not be fixed. The u-zarqa code point is used for the sifrei emet accent tsinnorit, and it can be used for stress-helper zarqa. Mam & uxlc use u-zarqa for stress-helper zarqa. (uxlc has only two such uses.) As a mnemonic, I think of u-zarqa as an accidentally truncated version of u-zarqa stress helper.

Although mam & uxlc use u-zarqa for stress-helper zarqa, I think it would be better to use u-zinor. That would be analogous to what is done for other accents with stress helpers: they use the same code point for both the stress helper & the true accent. Although no word in its corpus uses u-zinor for stress-helper zarqa, Taamey D supports this. This support is an exception to Taamey D’s usual policy of unconcern for words outside its corpus.

Using u-zarqa for stress-helper zarqa is like using u-qadma for stress-helper u-pashta, which is not done (nor should it be done) by mam or uxlc. (There is one word in the corpus that has both qadma & pashta though! This is a dually-cantillated word in the Decalogues. See the section “Qupo words” for information on dual cantillation.)

In the word [24] below, we can see some of the challenges associated with stress-helper zarqa. For each font, I show the standard u-zarqa/u-zinor (“zazi”) encoding, and then the u-zinor/u-zinor (“zizi”) encoding, which I prefer.

Taamey Frank & Sbl Hebrew handle these two encodings differently. For these two fonts, the “zizi” encoding causes the accent to be raised up above the ḥolam ḥaser dot, a somewhat radical way of mediating the competition for space between those marks. Though somewhat radical, one might prefer it to the crowding seen in these two fonts when the “zazi” encoding is used. And even this crowding still requires a someway right-biased (“early”) placement for the u-zarqa glyph. Taamey D’s glyphs for u-zarqa & u-zinor are a little narrower than those of Taamey Frank & Sbl Hebrew. This narrower shape, combined with a left-biased (“late”) placement of the ḥolam ḥaser dot, makes things come out a little better on a narrow letter like nun, in my opinion.

Note that Ezra Sil uses a different shape than the other fonts for u-zarqa & u-zinor. Ezra Sil mimics the font used in the influential bhs edition of the lc. Bhs, in turn, mimics the shape used in the lc for zarqa, tsinor, and tsinnorit.

Uxlc uses u-zarqa for more than just tsinnorit & stress-helper zarqa. It uses u-zarqa in ten cases where bhs puts a zarqa before the ascender of a final lamed. Bhs does this because that’s what the lc’s pointing scribe did in these cases. To me, this transcription is too literal. But as a font author, my job is to support every word in uxlc regardless of such judgments. Here’s one of the ten words:

וּֽלְיִשְׁמָעֵאל֘ [25]

Orphan sheva in yerushalayemah

In words for Jerusalem, the marks on or near the lamed should conform to the following pattern:

• Either pataḥ or qamats should appear under the lamed.
• Optionally, an accent can appear either above or below the lamed.
• An “orphan” vowel, ḥiriq or sheva, should appear in or near the space between the lamed & the next letter. (The next letter is always a mem of some kind.)

I call the vowel an orphan because its “parent” consonant, a yod, is only implicitly present, because each of these words for Jerusalem is a perpetual qere. If the orphan vowel is ḥiriq, the word then ends with a final mem. If the orphan vowel is sheva (rare), the word then ends with a mem and a he, with a qamats under the mem. All 4 of our standard fonts place the orphan ḥiriq far away from the true vowel under the lamed. (The true vowel under the lamed is pataḥ or qamats.) In fact, with default letter spacing, all 4 of our standard fonts place the orphan ḥiriq far enough the left to be under the final mem. Below, I show the pointed lamed & final mem of a Jerusalem word [26] with a geresh accent. I show these letters with default letter spacing and then with wide letter spacing below. The wide letter spacing shows that even though in default letter spacing, the orphan ḥiriq may appear to “belong” to the final mem, in fact it “belongs” to the lamed.

While all 4 fonts place orphan ḥiriq similarly, only Taamey D uses a position for the rare orphan sheva that is analogous to orphan ḥiriq, as in the word [27] below (only the suffix is shown):

The way the other 3 fonts place orphan sheva is not wrong, but it is not analogous to the way they place orphan ḥiriq. Note that if the orphan sheva is placed too far away from the true vowel under the lamed, the orphan sheva may start to look like the sheva-like part of a ḥataf qamats under the (non-final) mem. I think Taamey D strikes a good balance between these competing goals for the orphan sheva placement: notably far away from the true vowel under the lamed, but also not too close to the qamats under the (non-final) mem.

Above, the examples of orphan sheva with geresh were encoded in what I call laeo form: lamed, “ah” vowel, sheva, over-accent (above-accent). This is how uxlc encodes such words. The encoding mam uses is lao-e: lamed, “ah” vowel, over-accent, u-cgj, sheva. Below we can see that some fonts do not “like” lao-e:

Note that, possibly by accident, with lao-e, Taamey Frank looks almost the same as Taamey D. (Taamey Frank puts any marks after u-cgj at the left edge of the base letter, which is the default (zero) position for marks. In this case lamed is the base letter.) The fonts that struggle with lao-e do not struggle with its ḥiriq version, lao-i.

I wonder whether many of the challenges of orphan vowels might have been alleviated by having a u-implicit yod code point in Unicode. Then these orphan vowels would only be orphans in a visual sense. In terms of their encoding, these vowels would have u-implicit yod as their “parent” letter. u-implicit yod would typically render as not only invisible but also taking up no space.

Qupo words

There are 3 sections of the Hebrew Bible that have two systems of cantillation:

• Gen. 35:22 (in the Saga of Reuben)
• Exo. 20:2–13^M17^U (the Exodus Decalogue)
• Deut. 5:6–17^M21^U (the Deuteronomy Decalogue)

(Unfortunately, mam and uxlc have different verse numbering in the Decalogues. Although they agree on the number of the first verse, the number of the last verse is 4 higher in uxlc. Above and elsewere, we use the notation a^Mb^U to mean “verse a in mam and verse b in uxlc.”)

In Taamey D’s corpus, these sections notate both systems of cantillation on the same set of letters. As a result, these sections of the Hebrew Bible are the sections most densely packed with typographic challenges. Within these challenging sections, perhaps the most challenging words are the two qupo words, panay [28/29] & mitaḥat [30/31] These words contain letters that bear a burden of not only two accents but also two vowels! Qupo is acronym I use to describe these 4 marks:

• qamats
• under-accent (siluq or etnaḥta) (u-meteg used for siluq)
• pataḥ
• over-accent (revia or geresh)

Below are panay & mitaḥat in our standard 4 fonts.

It is perhaps not coincidental that Sbl Hebrew struggles with panay since until recently, an error in wlc made this word far less challenging. In other words, I suspect panay as shown above was not part of Sbl Hebrew’s corpus of concern.

The qupo words vary in appearance when rendered in Ezra Sil, depending on what software is used. In some software, the words look great. In other software, the below-marks still look great, but the above-accent is “late” (too far to the left).

The qupo words above are shown as they appear in mam, which uses qu-po encoding, where the dash indicates a u-cgj. Some fonts fare better with different encodings, but I think qu-po is the best font-neutral encoding.

Other dually-cantillated words

Normally, dually-cantillated words mixing an above-accent & a below-accent pose no great challenge since the two accents do not compete for space. But such above/below pairs can thwart other functions of the font. In the example [32/33] below, the logic to avoid collision between ḥolam ḥaser & zaqef qatan seems to have been thwarted, in Taamey Frank & Sbl Hebrew.

(As an aside, note that not only is rafeh present here, a rarity in Taamey D’s corpus, but also it is present with dagesh. This doubly-extraordinary notation indicates that in one cantillation, dagesh is present, but not in the other. This shows that cantillation, though primarily concerned with accents, can influence whether dagesh is used, and, as we saw with the qupo words, can influence whether pataḥ or qamats is used.)

Narrow letters like yod & vav are easily overburdened with below-marks, and, depending on context, may need some extra space before or after them. This is particularly true in the following example [34/35] from the Decalogues:

Overburdened letters

Above I showed a greatly-overburdened letter, vav, with 3 below-marks due to dual cantillation. But similar, albeit less extreme situations occur frequently even with the standard burden of a vowel & a below-accent:

(References are [36], [37], and [38].) Multiple below-marks are the most common cause of overburdened letters, but descenders can cause issues as well in some fonts. Below, Ezra Sil struggles a little, even with as wide a letter as kaf. (References are [39] & [40].)

Floating ḥolam ḥaser

In Taamey D, a ḥolam ḥaser dot sometimes floats forward onto the right arm of an alef, as in the words below:

(The affected syllables are highlighted and repeated with wide letter spacing.)

This float only happens when the alef is playing a vowel role, i.e. when the alef is a mater lectionis. When the alef is playing a consonant role, the ḥolam ḥaser dot does not float forward, as in the words below:

(The syllables starting with alef are highlighted. Those syllables and their preceding letters are repeated with wide letter spacing.)

Taamey D behaves like this in order to match the behavior of Taamey Frank, which likely inherited this behavior from Sbl Hebrew. Sbl Hebrew and Ezra Sil likely have this behavior in order to mimic bhs.

Notes on the limited corpus

Taamey D does not claim to be generally applicable. It supports many words outside its corpus, but this support is accidental. It often supports its specific corpus through general means, leading to this “accidental generality.” A concrete example is its support for letters with more than one above-accent. Excluding revia mugrash, such letters are rare in the Hebrew Bible (in mam there are about 20). Taamey D supports every such case in its corpus, and, accidentally, supports many cases outside its corpus. It supports those out-of-corpus cases because it was easier to support multiple accents in a somewhat general way than it would have been to add special-case support for only the in-corpus cases.

Since Taamey D is only accidentally general, it might not support words like the following:

• Taamey D might not support words from editions of the Hebrew Bible that do not closely follow uxlc or mam. For example, such words appear in mam’s documentation notes. (I do not include mam’s documentation notes in Taamey D’s corpus.)
• Taamey D might not support words with hypothetical cantillation. For example, Jacobson’s book on cantillation often shows different ways a verse might have been cantillated. Such a verse might contain words that are not in any edition of the Hebrew Bible.
• Taamey D might not support words from cantillated texts outside the Hebrew Bible. For example, here is a recent cantillation of an excerpt from Ben Sira.
• Taamey D might not support words from editions of the Hebrew Bible representing the Masorah parva. Those would likely need Masorah circle & combining dot above. Taamey D has glyphs for these code points but their placement is untested.

Texts pointed with only vowels & meteg (i.e. texts without cantillation) are unlikely to cause problems for Taamey D. So, a siddur is unlikely to cause problems for Taamey D. A siddur contains only a few small cantillated passages, and those passages are Biblical. Those passages are unlikely to have any words outside Taamey D’s corpus.

This document now concludes with the following:

• a section providing the book, chapter, and verse for each of this document’s numbered Biblical references
• an index of those Biblical references in order of their appearance in the Hebrew Bible

Numbered Biblical references

Index of Biblical references