(c) 2014-2021 Joost Kremers
Parsebib is an Elisp library for reading bibliographic database files. It supports both BibTeX / biblatex (.bib) files and CSL-JSON (.json) files.
The library provides functions that parse the current buffer. They are intended to be used inside a with-temp-buffer combined with insert-file-contents, but they can also be used in a buffer visiting a bibliography file, of course.
The bibliographic data is returned as a hash table. To parse multiple files, you can either insert them all into one temp buffer, or pass the hash table obtained by parsing the first buffer as an argument when parsing the next buffer.
The data in the bibliography file can be returned in two ways. The first option is for parsebib to return the contents of the file accurately. This means that the field values as returned by parsebib are literally the field values in the file. For various reasons, however, this representation is not ideal if you want to present the content of a bibliography file to the user with the aim of selecting one or more entries
Therefore, the second option is for parsebib to return the field values in such a way that they are suitable for display. For BibTeX / biblatex files, this means that @String abbreviations are expanded and cross-references are resolved. For CSL-JSON files, it means that field values that are not strings (notably name and date fields) are converted to strings in a sensible way.
Although both are bibliography file formats, there are obviously differences between BibTeX / biblatex on the one hand and CSL-JSON on the other. The entry types and field names are different, and CSL-JSON does not have something similar to @String abbreviations or cross-references.
Especially relevant for the purpose of this library is that there are differences in the format of the data returned for the two types of files. The bibliographic data is returned as a hash table. In this hash table, each entry is stored under its entry key (which is the id field in CSL-JSON) as an alist of (<field> . <value>) pairs. In BibTeX data, <field> is a string and field names are case-insensitive, so you may have "Author" or "author", and both may occur in a single .bib file. Furthermore, <value> is always a string.
In CSL-JSON data, the format of the alist is slightly different: <field> is not a string but a symbol and the symbol name is case-sensitive. The CSL-JSON standard describes which field names are lower case (most are) and which are upper case (ISBN, DOI, etc.) Furthermore, the <value> part of the alist items may be a string, a number or a vector, though when parsing a .json file, you can have parsebib convert all values to strings.
You can access the data in an entry's alist with assoc or alist-get, but in order to accommodate for the differences in the type of field names, you may want to use assoc-string instead, which generalises over symbols and strings and has a case-fold argument.
Another thing to note is that in BibTeX data, the type and key of an entry are stored in the entry's alist under =type= and =key=, while the same information is available in CSL-JSON data under type and id, respectively. While some of the core information on an entry is available in fields with the same name in both formats (esp. author, editor and title), most fields are named differently (i.e.., the year field in BibTeX corresponds to the issued field in CSL-JSON).
One last difference to note is that the general buffer-parsing functions, parsebib-parse-bib-buffer and parsebib-parse-json-buffer do not have the same type of return value. See the function descriptions below for details.
There is also a function parsebib-parse that takes a list of files and returns the entries in them. It has a slightly simplified interface, but it is format-agnostic: it handles both .bib and .json files.
Support for .bib files comes in two different APIs, a higher-level one that reads all items in one go, and a lower-level one that reads one item at a time. Parsebib supports @Preamble, @String and @Comment items, and obviously actual bibliographic entries.
In order to return entries in a way that is suitable for display, parsebib can expand @string abbreviations and resolve cross-references while reading the contents of a .bib file. When @string abbreviations are expanded, abbreviations in field values (or @string definitions) are replaced with their expansion. In addition, the braces or double quotes around field values are removed, and multiple spaces and newlines in sequence are reduced to a single space.
Resolving cross-references means that if an entry that has a crossref field, fields in the cross-referenced entry that are not already part of the cross-referencing entry are added to it. Both BibTeX's (rather simplistic) inheritance rule and BibLaTeX's more sophisticated inheritance schema are supported. It is also possible to specify a custom inheritance schema.
Expanding @Strings and resolving cross-references can also be done across files, if the result of parsing one file are passed as arguments when parsing the next file. Details are discussed below.
Note that if you wish to resolve cross-references, it is usually also necessary to expand @String abbreviations, because the crossref field may contain such an abbreviation. Resolving such a cross-reference will not work unless the abbreviation is expanded.
When parsing a bibliography file for display, one may not be interested in all the data of each entry. The higher-level API functions can therefore take a list of fields to be read and included in the results. Fields not in this list are ignored, except for the =key= and =type= fields, which are always included.
If you use this option and also want to resolve cross-references, you need to include the crossref field in the list of requested fields. Without it, parsebib is not able to determine which entries cross-reference another entry and no cross-references will be resolved. Also note that cross-referencing may add fields to an entry that are not on the list of requested fields. For example, in biblatex, the booktitle field of an InBook entry is linked to the title field of the cross-referenced Book entry. In such a case, if title is on the list of requested fields, the booktitle field is added to the cross-referencing entry, even if booktitle is not on the list of requested fields.
The higher-level API consists of functions that read and return all items of a specific type in the current buffer. They do not move point. Note that the arguments in these functions (except in parsebib-expand-xrefs) are keyword arguments.
Collect all entries in the current buffer and return them as a hash table, where the keys correspond to the BibTeX keys and the values are alists consisting of (<field> . <value>) pairs of the relevant entry. In this alist, the BibTeX key and the entry type are stored under =key= and =type=, respectively. Note that both <field> and <value> are strings.
The argument entries can be used to pass a (possibly non-empty) hash table in which the entries are stored. This can be used to combine multiple .bib files into a single hash table, or to update an existing hash table by rereading its .bib file.
If the argument strings is present, @string abbreviations are expanded. strings should be a hash table of @string definitions as returned by parsebib-collect-strings.
If the argument inheritance is present, cross-references among entries are resolved. It can be t, in which case the file-local or global value of bibtex-dialect is used to determine which inheritance schema is used. It can also be one of the symbols BibTeX or biblatex, or it can be a custom inheritance schema. Note that cross-references are resolved against the entries that appear in the buffer above the current entry, and also against the entries in the hash table entries.
The argument fields is a list of names of the fields that should be included in the entries returned. Fields not in this list are ignored (except =type= and =key=, which are always included). Note that the field names should be strings; comparison is case-insensitive.
Collect all @string definitions in the current buffer and return them as a hash table. The argument strings can be used to provide a hash table to store the definitions in. If it is nil, a new hash table is created.
The argument expand-strings is a boolean value. If non-nil, any abbreviations found in the string definitions are expanded against the @string definitions appearing earlier in the .bib file and against @string definitions in strings, if provided.
Collect all @preamble definitions in the current buffer and return them as a list.
Collect all @comments in the current buffer and return them as a list.
Find and return the BibTeX dialect for the current buffer. The BibTeX dialect is either BibTeX or biblatex and can be defined in a local-variable block at the end of the file.
Collect all BibTeX data in the current buffer. Return a five-element list:
(<entries> <strings> <preambles> <comments> <BibTeX dialect>)
The <entries> and <strings> are hash tables, <preambles> and <comments> are lists, <BibTeX dialect> is a symbol (either BibTeX or biblatex) or nil.
If the arguments entries and strings are present, they should be hash tables with equal as the :test function. They are then used to store the entries and strings, respectively.
The argument expand-strings functions as the same-name argument in parsebib-collect-strings, and the arguments inheritance and fields function as the same-name arguments in parsebib-collect-bib-entries.
Note that parsebib-parse-bib-buffer only makes one pass through the buffer. It is therefore a bit faster than calling all the parsebib-collect-* functions above in a row, since that would require making four passes through the buffer.
Expand cross-references in entries according to inheritance schema inheritance. entries should be a hash table as returned by parsebib-collect-bib-entries. Each entry with a crossref field is expanded as described above. The results are stored in the hash table entries again, the return value of this function is always nil.
This function can be useful if you use the lower-level API to parse .bib files, because in that case, resolving cross-references cannot be done while reading entries.
The lower-level API consists of functions that do the actual reading of a BibTeX item. Unlike the higher-level API, the functions here are dependent on the position of point. They are meant to be used in a while loop in which parsebib-find-next-item is used to move point to the next item and then use one of the parsebib-read-* functions to read the contents of the item.
All functions here take an optional position argument, which is the position in the buffer from which they should start reading. The default value is (point).
Find the first BibTeX item following pos, where an item is either a BibTeX entry, or a @Preamble, @String, or @Comment. This function returns the item's type as a string, i.e., either "preamble", "string", or "comment", or the entry type. Note that the @ is not part of the returned string. This function moves point into the correct position to start reading the actual contents of the item, which is done by one of the following functions.
These functions do what their names suggest: read one single item of the type specified. Each takes the pos argument just mentioned. In addition, parsebib-read-string and parsebib-read-entry take an extra argument, a hash table of @string definitions. When provided, abbreviations in the @string definitions or in field values are expanded. Note that parsebib-read-entry takes the entry type (as returned by parsebib-find-next-entry) as argument.
parsebib-read-entry also takes an optional argument keep-fields. This is a list of names of the fields that should be included in the entries returned. Fields not in this list are ignored (except for =type= and =key=, which are always included). Note that the field names should be strings; comparison is case-insensitive.
The reading functions return the contents of the item they read: parsebib-read-preamble and parsebib-read-comment return the text as a string. parsebib-read-string returns a cons cell of the form (<abbrev> . <string>), and parsebib-read-entry returns the entry as an alist of (<field> . <value>) pairs. One of these pairs contains the entry type =type=, and one contains the entry key. These have the keys "=key=" and "=type=", respectively.
Note that all parsebib-read* functions move point to the end of the entry.
The reading functions return nil if they do not find the element they should be reading at the line point is on. Point is nonetheless moved, however. Similarly, parsebib-read-entry returns nil if it finds no next entry, leaving point at the end of the buffer. Additionally, it will signal an error of type parsebib-entry-type-error if it finds something that it deems to be an invalid item name. What is considered to be a valid name is determined by the regexp parsebib-bibtex-identifier, which is set to "[^^\"@\\&$#%',={}() \t\n\f]*", meaning that any string not containing whitespace or any of the characters ^"@\&$#%',={}() is considered a valid identifier.
The support for CSL-JSON files comprises just one function: parsebib-parse-json-buffer. The actual parsing of the JSON data is performed by Emacs itself, either by the native JSON parsing routines (starting with Emacs 27.1, if available), or the built-in Elisp library json.el. Parsebib makes sure that the data is returned in a format that is similar to what is returned for .bib files.
When returning entries in a form that is suitable for display, the most important issue in CSL-JSON files is the fact that certain fields do not have string values. For example, name fields (author, editor, etc.) and date fields (issued, submitted etc.) are JSON arrays. Parsebib can convert these to strings if requested.
As with .bib files, it is possible to have parsebib only return specific fields when reading .json files. Here, too, the fields that identify an entry, i.e., id and type, are always included and do not need to be requested explicitly.
Collect all CSL-JSON data in the current buffer and return the result. The return value is a hash table, where the keys correspond to the identifiers of the entries and the values are alists consisting of (<field> . <value>) pairs of the relevant entry. In this alist, the identifier is stored under the key id and the entry type is stored under type. <field> is a symbol, while <value> can be a string, a vector (array) or another alist.
The argument entries can be used to pass a (possibly non-empty) hash table in which the entries are stored. This can be used to combine multiple .json files into a single hash table, or to update an existing hash table by rereading its .json file.
Some field values in CSL-JSON are not strings. These are primarily name and date fields, which in CSL-JSON are represented as JSON objects. The argument stringify determines how they are returned. When stringify is set to nil, they are returned as alists; with stringify set to t, they are converted to strings.
The argument year-only controls the way dates are converted to strings. If it non-nil, only the year part is returned. This argument only takes effect if stringify is set to t. See below for details.
The way values are converted to strings can be customised to some extent by the use of certain special variables, discussed below.
The argument fields is a list of names of the fields that should be included in the entries returned. Fields not in this list are ignored (except type and id, which are always included). Note that the field names should be symbols; comparison is case-sensitive.
Note that all arguments in this function are keyword arguments.
Convert the CSL-JSON data in entry to string form. entry is an alist as stored in the hash table returned by parsebib-parse-json-buffer. Return value is the modified entry.
Note that this function modifies entry in place. If you need to keep the original entry, call this function on a copy.
If the argument year-only is non-nil, date fields are shortened to just the year.
For details on the conversion, see below.
Convert the value of field to string form and return the result. field is a cons cell of the form (<field> . <value>), i.e., an element in the alist of an entry in the hash table returned by parsebib-parse-json-buffer.
The return value is a string representation of <value>. Note that unlike parsebib-stringify-json, this function does not modify its argument.
The following conversions are applied:
- Strings are returned unchanged.
- Numbers are converted to strings using
(format "%s" number). - Name fields (as defined by
parsebib--json-name-fields) are converted usingparsebib-json-name-field-templateandparsebib-json-name-field-separator; see below for details. - Date fields (as defined by
parsebib--json-date-fields) are converted to a format2021-4-22. If only a year is present, the month and day parts are omitted. Theseasonandcircafields are accounted for, and so areliteralandraw. - Fields with an array as value (currently, in v1.0 of CSL-JSON, this only applies to the
categoriesfield), are converted to a string usingparsebib-json-field-separator; see below for details. - Anything that doesn't match any of the categories above is converted to a string using
(format "%s" value), after which any newlines are removed and replaced with a space. This is a catch-all that shouldn't be necessary in valid CSL-JSON files.
The optional argument short only applies to date fields. If short is non-nil, a date field contains just the year; month and day parts are ignored. If no year part can be found, short returns the string XXXX. Note that with short, other parts of the date field are ignored.
This variable holds the template used for converting name fields. Its default value is:
"{non-dropping-particle }{family, }{given}{ dropping-particle}{, suffix}{literal}"
The elements of a name field appear in the string in the order provided in the template. Elements that are not part of a particular name field are ignored, including any punctuation and white space contained inside the braces. So for example, if a name field does not have a suffix, the comma-space appearing inside the {, suffix} part of the template is not included in the final string.
It is possible to modify the form of the resulting string by let-binding parsebib-json-name-field-template before calling parsebib-parse-json-buffer, parsebib-stringify-json or parsebib-stringify-json-field.
If more than one name appears in a name field, they are separated by the value of this variable in the string. The default value is " and " (note the spaces). It is possible to let-bind this variable.
Field values that are arrays are converted to a string using the value of this variable as a separator. Currently (CSL-JSON v1.0), this only applies to the categories field, which is an array of strings. The default value of this variable is ", " (note the space). It can be let-bound like the variables above.
Parse a bibliography file or list of files and return the entries in them. This function can be used for both .bib and for .json files, and also for a combination thereof. It returns all entries in all files in a single hash table.
This is a high-level function meant for retrieving bibliographic entries in such a way that they can be shown to a user. It is not possible to retrieve the @Preamble or @Comments in a .bib file using this function. Use parsebib-parse-bib-buffer or one of the other functions for that.
parsebib-parse basically just calls parsebib-parse-bib-buffer or parsebib-parse-json-buffer as appropriate and passes its arguments on to those functions. The argument entries is passed to both, as is fields. The field names in fields need to be strings, regardless of the file format, though. parsebib-parse converts the strings to symbols when it parses a .json file. The strings argument is only passed to parsebib-parse-bib-buffer, since there are obviously no @Strings in a .json file.
The display argument controls the way in which the entry data is returned. By default, it returns the data in a way that is suitable for display. For .bib files, this means that @String abbreviations are expanded and cross-references are resolved. For .json files, it means that fields are returned as strings and that month and day parts in date fields are ignored.
See the doc strings of parsebib-parse, parsebib-parse-bib-buffer and parsebib-parse-json-buffer for details on the meaning of display.