Article about naming files in a meaningful way. Naturally files should have unique names so we don't end up with several files with the same or very similar names. This article takes file naming one step further by looking at how the file name itself can carry useful information about the file.
Why meaningful filenames
If all names are kept unique why also try to embed meaning in the file name itself? Here are some approaches to managing information about information which which you might recognise:
- File names and directory hierarchies can help describe the contents. By placing a picture called second birthday.jpg in a directory called My son James we have stored data (the picture relates to James' second birthday) about the data (the picture of the birthday party).
- Data about data (called metadata, (Wikipedia's Metadata entry) can also be stored inside the file it describes, for example:
- HTML, the language of webpages, uses tags like Example. Here the meta data describes the style of the text, ie: Example is normalText
- EXIF (Wikipedia's EXIF entry) is a way of storing meta data in image files, like when the photo was taken and what type of camera was used.
- Database systems (Gramps is a database system for genealogy) can store a huge amount of data about data. They are very efficient at this job and very powerful.
- Google Search uses a database to remember what web pages are about, and tells you when you ask.
So why not use one of those options?
- EXIF is great, but only for some types of files (not supported in JPEG 2000, PNG, or GIF), there are lots of different systems for different types of files. People are working hard to improve this situation all the time.
- HTML is great if you can store all your information as HTML files, but HTML files cannot contain other files, they just point to them. So we'd basically end up making a website about our files.
- A database, well we already use this when we use Gramps. The Gramps database stores lots of information about the files and records it records. But Gramps does not store the actual file inside the database. If the connection between Gramps and the data it is describing is broken, then the files are just files. They contain no more information than they did when you first imported them into Gramps.
This system of meaningful filenames has the following aims:
- Preserving enough metadata to give the file's content context without Gramps
- Creating file names normal people can understand so they can see what the file is about without Gramps
- Creating file names which a computer can process easily so files need to be batch processed and metadata can be read directly from the file name without possible confusion
- Creating a system simple enough to use all the time for every file
To be understandable we need to be able to use full words where appropriate.
To be computer readable we need to separate the parts in a way which a script can easily recognise and, more importantly, in a way which would never occur in real language. So it would be no good to mark a name section with the word name if we also can use the word name somewhere in the file where it is not meant to be a marker.
To be simple enough to remember the system should not be too complicated, after all Gramps is meant to store the real information, this is just a supplement.
What's in a name?
It would be nice if we could have files called
Marriage of Mary Angus Jones and Matthew Williams, 2nd Dec 1923 (William Angus is to Mary's right).jpg
But this meets only one of the criteria above, that of understandable filenames. How can a computer know who got married? what their surnames are? and so on. And anyway because of the limitations of Portable Filenames we can't have file names like that. We have to drop the reliance on capitalisation, drop the spaces, drop the comma and drop the brackets. To be computer readable we need to separate the sections with a system of markers to indicate where the surname, event name etc are.
So what sections do we want to be able to identify? Here's a basic list that should be enough for most situation, remember that Gramps stores the more complex information, we're just trying to give a useful structure to our files.
- Event type
Some more important criteria. All file names:
- Must be unique
- Must have all necessary information
- Must have no more information than necessary
So if I find a file somewhere strange in my system, or if someone I sent a file to seven years ago says "that file you sent me - that's not Jean it's her daughter" I know where my archive copy of that file will be.
This is a system contributed by Duncan Lithgow.
If we base a naming system on the 3 and 4 letter Lineage-Linked GEDCOM Tag Definition used in the GEDCOM 5.5 standard we have a good long list of tags to chose from. By limiting the GEDCOM tags list we can make the following shortlist (which does not include events):
AUTH-- Author "The name of the individual who created or compiled information." DATE-- Date EVEN-- Event "A noteworthy happening related to an individual, a group, or an organization." GIVN-- Given name "A given or earned name used for official identification of a person." NAME-- Name, use only if GIVN and SURN are not known "A word or combination of words used to help identify an individual, title, or other item. More than one NAME line should be used for people who were known by multiple names." NOTE-- Note "Additional information provided by the submitter for understanding the enclosing data." PLAC-- Place "A jurisdictional name to identify the place or location of an event." REFN-- Reference "A description or number used to identify an item for filing, storage, or other reference purposes." SOUR-- Source "The initial or original material from which information was obtained." SURN-- Surname "A family name passed on or used by members of a family." TITL-- Title "A description of a specific writing or other work, such as the title of a book when used in a source context, or a formal designation used by an individual in connection with positions of royalty or other social status, such as Grand Duke."
Each marker ends with two hyphens (--). Two because we can't rely on the marker being recognised as capitalised, so a surname like Besour-Jean could be mistaken for beSOUR-Jean and the system thinks that SOUR- marks a source section.
In order for the file name to be parsed as meaningful text I think some we also would need
_ Underscore to represent a space __ Double underscore to represent a comma followed by a space
The GEDCOM 5.5 standard defines so few events as to be useless. The Gramps XML schema defines no events as these can be made by the user. This all seems fair enough since events are highly culture based. The situations where I think a set of events should be defined are those which will be connected with source records. GEDCOM has a reasonable group of those but they are heavily based in western christian culture. The solution must be language and culture dependent. Here's my list:
marriage is for an actual marriage event and all the associated documentation, including possible divorce and separation documentation. birth is for the actual birth records, also christening record death is for death records census is for census records civic is for military service records, and government records of any type health is for health records
An event image file
This could be parsed (by Gramps?) as the description:
Event: Marriage Surname: Jones Given name: Mary-jean Surname: Williams Given name: Matthew Date: 2nd Jan, 1923 Note: William angus to the right of mary
or it could make the text:
Mary-jean Jones and Matthew Williams, marriage 2nd Jan 1923. (William angus to the right of mary)
A source image file
This could be parsed (by Gramps?) as the description:
Source: Census Place: Uk, england, london Date: 21st March, 1840 Surname: Jones Given name: Mary-jean
or it could make the text:
Census, Place: Uk, england, london, on 21st March 1840. This is a source connected to Mary-jean Jones
A source text
This could be read as the description:
Source: Publication Title: The Jones Family from 1735 Author: Mary Jean Jones Date: 1872
Or it could make the text:
"The Jones Family from 1735" by Mary Jean Jones, 1872
Over at Wikipedia there is a good explanation of a SWOT analysis.
|File length||Holds a lot of information||All the information is already in the genealogy software||Easily recognised. Easy to search for files with a certain Tag||?|
Gramps ID based
This is another attempt by Duncan Lithgow to find a good system. It is not finished so feel free to add comments and correct any obvious mistakes.
Here's the records we'll use as examples. They involve Mary Agnes Williams (daughter of John Williams and Anna Matthews). She married Anders Sørensen (son of Anders Sørensen and Anna ?) and they had a daughter Anna Sorensen, note the spelling change.
- Census record: mentioning her and her siblings and parents. It is from the 1810 census in the London parish of Dangerfield on Saint John Road.
- Portrait: a hand drawn portrait of Mary, undated, assumed to be from before her marriage.
- House picture: her parent's Saint John's Road row house in London, from some time around 1810's
- Court record: Anders Sørensen was before the district court for drunk and unbecoming behaviour on January 3rd, 1820. Engelfield, London.
- Marriage certificate: She married Anders Sørensen, 2nd December 1823, in London.
- Wedding portrait: in the picture is Anders Sørensen's father, also called Anders Sørensen (on the back it says that Anders Sørensen's (the son) mother is called Anna).
- Birth certificate: of Anna Sorensen (daughter of Mary and Anders) dated January 18th, 1824
- Family tree: a hand written family tree called "The Dean family from 1735" by an Angus Dean written in 1972 which connects the families Dean and Williams.
Please update or expand this section.
This system tries to meet the following aims:
- simple enough to remember
- just enough information, and no more
- all media for one family name is under one directory (portability for travel)
- all media for generating reports is under one directory (for portability)
The record types tell us what the record is about. Gramps ID's use the first character to denote the type of item the ID refers to. Sticking to something already thought and taking the most relevant ones to stored records these can be used as the following tags for record types:
- I-- Individual
- P-- Place
- E-- Event
- S-- Source
(see also source types)
- Records about repositories?
- Correspondence with family?
- What about records covering more than one type?
- What will happen on old 8+3 file systems?
Properties tell us just enough information to make the file name meaningful and recognisable, and split this information up so we can search for parts of it with our file manager. It's the what, where, when, why and how of what's in the record.
By making all properties of each record compulsory we avoid extra tags like GN for given name and so on. We can see what a property is by where it is in the file name.
- family name is their surname before marriage, but including deed pool changes, MacArthur for example
- given name is their official first name
- uid is a unique identity, in this example the (original) Gramps ID of the media file
- source date is the date in ISO 8601 format when the information left the people or organisation responsible for it
- event date is the date in ISO 8601 format when the event occurred or started. YYYY-MM-DD, ie. 2008-12-28
- event type is a noun describing the event, chosen from a list of event types, ie: marriage
- title is the name of a document (book, letter, census) or object (gravestone, heirloom), ie. williams__arthur_headstone
- source author name is the name of the person or organisation most responsible for the information. For people always use family name first followed by two underscores (__), ie: church_of_lds
- note is for notes. Names should always be family name first followed by a double underscore
Now we can outline a single schema for all record types in which the following rules apply.
- File names are written directly to the file name, not copied from another program.
- File names start with a single capital letter representing their record type.
- Record properties are separated by two dashes (--). This can not be used for anything else.
- Missing information is replaced by a single underscore (_).
- Names in notes should always be family name first and separated by two underscores, ie: doe__john which can be represented as John Doe or Doe, John.
- Place names should start with the largest geographical region followed by a double underscore before the next geographical region, ie: oz__far_far_away__yellow_brick_road which can be represented as Oz, Far far away, Yellow brick road.
- If the family name is unknown it must be replaced by an underscore. This will give three consecutive underscores (___), ie: ___john' should always be interpreted as meaning [no record], John.
- event types should always be drawn from a list to avoid separate words being used for the same event type. (Maybe use the event list gramps uses?)
- <record type>-- (I, P, E or S)
- <source type/event type>-- (needs expansion.)
- <1st persons family name[__2nd persons family name]>-- (two names for couples or families, alphabetical)
- <1st persons given name(s)[__2nd persons given name(s)]>-- (two names for couples, same order as for family names)
- <country code__region__city>-- (use as many divisions as needed)
- <date>-- (ISO date, YYYY-MM-DD)
- <note>-- (usually not needed)
- <uid> (a Unique ID, possible derived from the gramps ID)
Here's a version of the naming structure for quick reference.
Using the records outlined in the beginning we would get the following file names. (Please help complete this list of examples)
- Census record: S--census--matthews__williams--anna__john--uk__london__dangerfield__st_johns_rd--1810-_-_--_-00874.pdf
- Court record: S--court_record--soerensen--anders--uk__london__engelfield--1820-01-03--before_district_court--00826.pdf
- Marriage certificate: S--marriage_certificate--jensen__williams--anders__mary_agnes--uk__london--1823-12-02--_--00864.pdf
- Portrait: I--portrait--williams--mary_agnes--uk__london--1823-12-03--wedding_portrait--000967.jpg
One possible shortcoming of using an event and/or individual based naming strategy is that it could "clash" with the relation between a filename and the source. This only applies to filenames that are a representation of a specific part of a source.
An example: having found the baptism record of Anna in page 51v of the 1843-1850 Baptism book of a certain Church we save the image (which displays pages 50v and 51f, i.e. it is an image of the "open book") and name it something like BAP--Anna--1850.png (just an example, any individual and role based mechanism will yield similar results). This works fine and allows one to easily extrapolate information from the file name.
However, we latter find that in the exact same page, but a few paragraphs below, we have the baptism record of another individual, from another part of the family tree. While we can simply use the original file it wouldn't convey the right information. We can duplicate the file, but that doesn't make much sense, especially since when adding the file to the Source gallery we would end up with a duplicate, which makes little sense.
One way to deal with this is to use a purely source-based approach in naming the files. The downside is that event and individual information can't be gleaned by looking at the file name - one would have to use Gramps itself to maintain the appropriate relations, which is after all something that is part of the source referencing work that should be done. On the other hand, and when talking about Sources that are books, it allows for easy grouping of content related to the same source, e.g. all the relevant pages on a certain book. An example filename would be "PBL--BAP3--F51-52.png", where PBL is the short name of the source author and BAP3 the short name of the specific source. Longer, more descriptive filenames could be used by using full names instead of codes.
Obviously this method is limited in scope to some kinds of sources, and doesn't make sense for naming photos or documents that aren't part of a larger source (e.g. an ID card).
A really good article on why and how to implement this kind of naming system is Hierarchical Sources by Tony Proctor.