MathType MTEF v.5 (MathType 4.0 and later)

MTEF data exists in the following versions:

Version 2 differs from version 1 only in the format of the header.

The MTEF version 5 header contains:

MTEF data consists of a series of records. Each record starts with a record type byte and an options byte, then is followed by data specific to the record.

The overall structure of an MTEF stream is:

Once the header has been read, reading of the MTEF stream should be driven by reading the next record type and then acting on it. If a given record depends on data that is not contained within the record itself, such data is defined by records that precede it.

The following record types are used:

If the record type is 100 or greater, it represents a record that will be defined in a future version of MTEF. For now, readers can assume that an unsigned integer follows the record type and is the number of bytes following it in the record (i.e. it doesn't include the record type and length). This makes it easy for software that reads MTEF to skip these records. Although it might be handy if all records had such a length value, it will only be present on future expansion records (i.e. those with record types ≥ 100).

LINE, CHAR, TMPL, PILE, MATRIX, and RULER records are followed by object lists that define contents of each equation structure. Each object list contains a sequence of records of any type and terminated by an END record. In a special case for LINE records, if there are no objects in the list, the line record will have the mtefOPT_LINE_NULL option set, in which case the object list is omitted entirely (i.e. no END record). Although there are no restrictions made by the MTEF format on what record types may occur in any particular list, the user interface prevents certain things from happening. For example, the object list defining the contents of a pile contains only LINE records.

Some of MTEF's records (FONT_STYLE_DEF, FONT_DEF, COLOR_DEF, and ENCODING_DEF) do not represent equation structure themselves but are referred to by equation structure records. The definition records in an MTEF stream are assigned indices, starting at 1 (except for ENCODING_DEFs, which start at 5), in the order they appear in the stream. Records which refer to these definitions do so using the index value of the definition. Definition records always appear in the stream before their first reference.

Each MTEF 5 record starts with a type byte followed by an option byte. This is different from earlier versions of MTEF where the option flags were stored in the upper 4 bits of the type byte.

The option flag values are record-dependent:

All dimensional values are expressed in MathType internal units, 32nds of a printer's point (a point is 1/72 inch).

Signed values are written as follows:

Unsigned values are written as follows:

Some values are written as 16-bits even if the value would fit in a byte. In this case, the value is written low byte followed by high byte.

LINE, CHAR, TMPL, PILE, MATRIX, and EMBELL records may store the result of nudging (small offsets applied by the user). A nudged record has the mtefOPT_NUDGE option (0×8) and the option byte is followed immediately by the nudge offset. The nudge offset consists of either two bytes or six, depending on the amount of offset. If -128 ≤ dx < +128 and -128 ≤ dy < +128, then the offsets are stored as two bytes, dx followed by dy, where each value has 128 added to it before it is written. Otherwise, two bytes of 128 are stored, followed by the offsets, dx and dy, stored as 16-bit values, low byte followed by high byte.

CHAR records contain a typeface value (biased by 128), written as a signed integer. If the value is positive, it represents one of MathType styles:

If the value is negative, it represents an explicit font as specified by a FONT record.

Typesize values (sometimes referred to as lsizes) are used in several MTEF records. Not all values may be valid in a particular record. Their meaning is as follows:

Character styles are represented by a single byte, bit 0 indicates bold and bit 1 indicates italic. In other words:

Horizontal alignment values are used in several MTEF records. Not all values may be valid in a particular record. Their meaning is as follows:

Vertical alignment values are used in several MTEF records. Not all values may be valid in a particular record. Their meaning is as follows:

In the EQN_PREFS record, sizes and spacing values are both written as dimension arrays. These arrays are used to record the particular settings from the Define Sizes and Define Spacing dialogs used to define the equation. Instead of recording each value as a number, the dimension array captures each value as a character string. This ensures the user always sees the value just as entered, rather than possibly modified by rounding and/or truncation in conversion (e.g. so "2.0 inches" doesn't turn into "1.999 inches").

Each array is written as a single byte count of dimensions in the array, followed by a "nibble stream" containing the dimensions. Each nibble is a 4-bit value, packed 2 per byte. The upper 4 bits of each byte precedes the lower 4 bits. If the array contains an odd number of nibbles, an additional 0 nibble is written to round out the stream to whole bytes.

Each dimension in the array consists of a units nibble (see first table below), followed by a nibble for each character in the value string (see second table below), terminated by a 0×F nibble. The following table shows how each units nibble is interpreted:

The following table shows how each nibble in a value string is interpreted:

Consists of:

There is no option byte.

Consists of:

The line spacing value, if present, is the distance between the baseline of this line and the line above it.

Consists of:

The character value itself is represented by one or more values. The presence or absence of these value is indicated by options and appear in this order:

The MTCode value defines the character independent of its font. MTCode is a superset of Unicode and is described in MTCode Encoding Tables. The 8-bit and 16-bit font positions are mutually exclusive but may both be absent. This is the position of the character within its font. Some of the common font encodings are given in Font Encoding Tables.

Consists of:

The template selector and variation codes determine the class of the template and various properties of the template, such as which subobjects can be deleted by the user (see Templates). The class of a template determines the order and meaning of each of its subobjects (see Template subobject order).

The variation code may be 1 or 2 bytes long. If the first byte value has the high bit set (0×80), the next byte is read and combined with the first according to this formula:

variation code = (byte1 & 0×7F) | (byte2 << 8)

The template-specific options field is only used for integrals and fence templates:

Consists of:

Consists of:

The values for valign, h_just, and v_just are described in PILE above.

The row partition line type list consists of two-bit values for each possible partition line (one more than the number of rows), rounded out to the nearest byte. Each value determines the line style of the corresponding partition line (0 for none, 1 for solid, 2 for dashed, or 3 for dotted). Similarly for the column partition lines.

Consists of:

The embellishment types are:

Consists of:

Each tab stop is described by a tab-stop type (0 for left, 1 for center, 2 for right, 3 for equal, 4 for decimal), followed by a 16-bit integer offset from the left end of the slot or pile with which it is associated.

Consists of:

This record associates a character style with a font. See Definition records.

Consists of one of the following cases:

if lsize < 0 (explicit point size):

else: (large delta)

Sizes in MathType are represented as a pair of values, lsize and dsize. Lsize stands for "logical size", dsize for "delta size". If it is negative, it is an explicit point size (in 32nds of a point) negated and dsize is ignored. Otherwise, lsize is a typesize value and dsize is a delta from that size:

Simple typesizes, without a delta value, are written using the records described in the next section.

Consists of:

These records are just short ways of specifying a simple typesize where dsize is zero. The tag value represents an lsize + 10. So if the tag value is 10, it means equation content following it will be Full size (szFULL), tag value 11 means szSUB, and so on. See typesize.

Consists of:

The appearance of this record in the stream indicates that all following equation records (until the next COLOR record) have the color defined by the indicated COLOR_DEF record.

Consists of:

This record defines a color (see Definition records). Each color value is written as a 16-bit integer that ranges between 0 and 1000 where 0 is the absence of the color and 1000 is a fully saturated color. So, an RGB color definition for black has all three components at 0.

Consists of:

This record associates an font encoding with a font name. See Definition records.

Consists of:

When reading arrays, the number of values may be less than or greater than expected. MTEF readers should be driven by the array count. If the array is shorter than expected, assume defaults for the missing values. If the array is longer than expected, the extra values must be skipped to stay in sync with the MTEF stream.

Spacing values are written in the following order:

The style definition array is written as a single byte count followed by that number of style definitions. The order is defined by Typeface values, however only fnTEXT through fnTEXT_FE are written. Each style definition is written as an unsigned integer that is 0 if the style is unused in the equation or is the index of the corresponding FONT_DEF record. If the style is used (not 0), it is followed by a single byte character style.

Consists of:

This record defines (see Definition records) a font encoding and is referred to by a FONT_DEF record. In order to reduce the size of the MTEF stream, the following 4 encodings are predefined:

This means that the first ENCODING_DEF record in the MTEF stream is considered to have an index of 5. See Extending MathType's font and character information and MathType's character encodings for more information on font encodings.

The following variation codes apply to all templates whose class is BigOpBoxClass or LimBoxClass:

Template subobjects appear in object lists in the order that matches the movement of the insertion point within a template in the MathType user interface. For template classes that have more than one slot, the following list shows the order of subobjects for all templates with that class:

ArroBoxClass: expanding arrows

BigOpBoxClass: integrals, summations, products, etc.

DiracBoxClass: three-part bracket structure

FracBoxClass: fractions with a horizontal bar

HFenceBoxClass horizontal expanding braces

LDivBoxClass long division

LimBoxClass limits

ParBoxClass parenthesized or bracketed slots

RootBoxClass radical sign

ScrBoxClass subscripts and superscripts

SlashBoxClass fractions with a diagonal bar