open list of unexpected changes #11
Comments
|
Updated after testing our current random sample set. Not too terrible and not all bugs. |
|
Following #12 all of the mathops tests are broken (extremely large output). |
WARNING includes new issues, see #11 (comment) WARNING squash before merge!
I noticed that there are cuberoot and fourthroot - @davidmjones should we map those back to traditional \sqrt[3] \sqrt[4] or are they meant to not have stretchy overlines? |
|
Fun find while creating a test file for this mathjax/MathJax#2678 |
|
jmath: Without unicode-math extension: With unicode-math extension: Note the space before the jmath when the extension is used. |
|
Re jmath: this probably comes down to MathJax using u0237 https://github.com/mathjax/MathJax-src/blob/29cd0e8569788ec41dd4b2198018983493b83e36/ts/input/tex/base/BaseMappings.ts#L118 and unicode-math using u1D6A5. The space between the preceding comma and jmath vs preceding comma and mu seem to match (and match CM rendering on mathjax.org#demo) so I'm no longer sure if the change introduced by unicode-math is bad or actually an improvement. @davidmjones what do you think? For reference, CM rendering on mathjax.org#demo |
With the latest STIX2, they are now both large. @davidmjones based on your comment at stipub/stixfonts#179 (comment) I think this is the expected change, correct? |
|
Updated the previous comment to clarify a bvit. But I'm wondering if they should both map to u2216 now. |
|
varnothing. This seems to come down to MathJax using with variantForm:true without: I think without is the correct form - @davidmjones wdyt? |
|
On integrals. We set moveablelimits on all of the following mathjax-unicode-math/ts/unicode-math.ts Lines 1534 to 1596 in cbca2de IIRC integrals are generally not moveable (cf. https://github.com/mathjax/MathJax-src/blob/29cd0e8569788ec41dd4b2198018983493b83e36/ts/input/tex/base/BaseMappings.ts#L153-L184) @davidmjones is that correct? are the any others the linked section that we should reconsider? |
|
Concerning |
|
Thanks, Davide. I'll leave that to David to decide. |
|
On "varprojlim arrow tighter and thinner (worse?)" This is actually about underleftarrow. The difference is that MathJax uses u2190 for underleftarrow cf. https://github.com/mathjax/MathJax-src/blob/29cd0e8569788ec41dd4b2198018983493b83e36/ts/input/tex/base/BaseMappings.ts#L520 while unicode-mathjax uses u20EE. With u2190:
With u20EE:
@davidmjones what do you think? If we decide we don't like the current mathjax-unicode-math output, then the question is which other arrows need checking and changing. |
|
Note that U+20EE is not in the operator dictionary in Appendix C of the MathML spec, so it doesn't get the |
|
On nabla being italic. Simplest answer: MathJax sets it to normal - alongside other symbols https://github.com/mathjax/MathJax-src/blob/29cd0e8569788ec41dd4b2198018983493b83e36/ts/input/tex/base/BaseMappings.ts#L113-L144. @davidmjones is unicode-math's behavior intentional? (If not, are there other characters we need to consider?) |
|
On \prime. MathJax has @davidmjones besides the 6 other macros containing "prime" are there others we'd need to consider? |
Just to clarify: the large version of the prime is considered to be the variant form, and the small-raised version is the normal form. That allows U+2032 and similar characters to be used as pseudo scripts properly. When used as |
|
Thanks for both comments, @dpvc! |
|
@davidmjones I've updated the OP with links to the various comments to make it easier to go through this. I suppose it's best to go through it F2F; then I'd create issues for the various pieces. |
No, it's not intentional. The \nabla should be upright. |
unicode-math.sty just defines them to produce the Unicode characters; i.e., leave them as-is. |
WARNING includes new issues, see #11 (comment) WARNING squash before merge!
This definition would break stretchy root overlines. Part of #11
Document this intentional change (which is an improvement). See also the discussion in #11.
|
Re underleftarrow (varprojlim in the top card): As per F2F, the problem stems from the fact that, in LaTeX both varprojlim and underleftarrow are defined using \leftarrowfill. However, MathJax uses u2190 for underleftarrow as it makes the most sense given its internal format. So we will follow MathJax here which in this case simply means disabling underleftarrow (and revisit related macros). |
Avoids problems in superscripts by falling back to core MathJax. Part of #11
|
For completeness, regarding \prime. Xetex+unicode-math is smarter with primes and keeps \prime (and others) in their normal ("superscripted") form while automatically applying ssty1 when primes are in a superscript. For now, we disabled / fallback for \prime to core MathJax and punt on the other primes. |
Set emptyset to use u2205's variant form (cf varnothing). NOTE This is STIX Two specific; CM has u2205 forms reversed. Part of #11
Disables many to fallback to core MathJax, and adjusts others. The main theme is to avoid combining above/below characters. Part of #11
From wider testing, let's collect unexpected changes
$$\begin{equation*} f_{k,m}(z) = \frac{1}{2\pi i} \oint _C \dfrac{\Delta ^\ell (z) E_{k'}(z) E_{14-k'}(\tau )}{\Delta ^{1+\ell }(\tau ) (j(\tau ) - j(z))} r^{-1-m} \, dr, \end{equation*}$$$$\begin{equation*} f_{k,m}(z) = \int _{-\frac{1}{2}+iA}^{\frac{1}{2}+iA} \frac{\Delta (z)^\ell }{\Delta (\tau )^{1+\ell }} \frac{E_{k'}(z)E_{14-k'}(\tau )}{j(\tau )-j(z)} e^{-2\pi i m \tau } \, d\tau . \end{equation*}$$$\sqrt {3}}$$\sqrt[3]{p}$$\partial W_\xi = {\widetilde{W}}_\xi \smallsetminus W_\xi$$y \in e_nJe_n \setminus e_n J^{n+1} e_n$parametrizes those $(Y,\jmath ,\mu )$ with $(Y[p^\infty ], \jmath [p^\infty ],\mu [p^\infty ]) \cong (X[p^\infty ],\iota [p^\infty ],\lambda [p^\infty ])$,$T = \varprojlim _{\times p} A[p^n]$$\nabla (i)$$\sigma (z,z^{\prime })=Jz\cdot z^{\prime }$$B_\varnothing \to B_\varnothing (2m)$The text was updated successfully, but these errors were encountered: