Information in its raw forms — databases, exports, variables, snapshots. The substance the systems move.
The gathering of many into one summary figure. Aggregation always hides the distribution inside it.
Short form of annotation, and of approximate nearest neighbour. A named abbreviation rather than a concept.
To add at the end without altering what came before. Appending is how a record grows while staying a record.
To place elements in deliberate relation — in time (a sequence), in space (a composition). The conductor's verb.
An ordered set reached by position. Also a display arranged for viewing — both are things laid out in order.
A tree showing a text's grammatical structure rather than its characters. What a parser produces once it has decided.
A compound fragment from this library's own naming, not an English word. Flagged for the mis-cut band rather than defined.
A named property belonging to a thing. Also to credit a thing to its source — the same word for a quality and for a claim of origin.
To fill in what was missed after the fact. Backfilling repairs a record's gaps and should mark that it did.
A group processed together as one run. Batching trades responsiveness for throughput.
Having exactly two states. Binary is a property of some systems and is imposed on many things that do not have it.
To join parts into one keepable whole — pages into a book, tracks into an album. Combination that produces an object.
A held run of bytes with no structure the system reads. A blob is content the store agrees not to understand.
A byte-order mark — an invisible sign at the start of a file saying how it is encoded. Small, unseen, and capable of breaking everything downstream.
A nearby copy kept so the far original need not be fetched again. A cache trades freshness for speed.
To take hold of something passing so it can be kept. Capture fixes what would otherwise be gone.
The smallest bounded unit of a living body or a grid. A cell has a wall, which is what makes it one thing.
Short form of characters — the single written signs a text is made of. A named abbreviation rather than a concept.
Those descended from or contained within another. Named from the parent's side, which is the whole of the relation.
A named store for meaning-based retrieval rather than exact matching. A named tool rather than a concept.
Cofra (Irish cófra: a chest, a coffer). A zip archive read and written with no dependency at all, which is to say a box with the list of what is in it fastened inside the lid. Entries are STORED or DEFLATE and nothing else. CRC-32 is computed over every entry AS IT IS READ and checked against the directory, and a mismatch names the entry and both values rather than returning a shorter file and calling it success. Zip64, encryption and every other compression method are REFUSED BY ENTRY NAME — method 12 is named bzip2, 99 is named AES, and a method the table does not know is refused by its number rather than guessed at — because a silent skip is the one failure a container must never have. An entry with no date takes the DOS epoch, 1980-01-01, the earliest an MS-DOS date field holds, which is the whole reason the same content written twice is the same bytes twice: nothing here ever asks the clock. The reader finds the end record by scanning back for its signature and confirming its comment length lands exactly on the end of the file; it reads the central directory rather than trusting the local headers' sizes, because flag bit 3 may defer them; it slices a stored entry rather than viewing it, so one small file cannot hold a whole archive in memory behind it; a trailing slash over zero bytes is a folder marker and is skipped; and a repeated name is refused in both directions rather than resolved. THE PLATFORM PRIMITIVE IS NEVER IMPLEMENTED: the host surface names raw deflate in both directions and nothing else — no level, no strategy, no dictionary, because the moment a host may be tuned two hosts stop writing the same bytes — and two reference hosts on two unrelated machines, the platform's own CompressionStream and node:zlib, are proven byte-for-byte identical rather than promised. No keys, no passwords, no ciphers, and no verb on the surface that could hold one. No disk, no clock, no network, zero runtime dependencies.
An upright division, in a structure or a table. Both a support that bears load and a field that holds one kind of thing.
Making a thing smaller while keeping what matters. Every compression decides what mattered.
A misspelling of context as it stands in this library's own names. Recorded as found rather than silently corrected.
To change a thing from one form into another while keeping what it holds. Conversion preserves content and not always fidelity.
A second thing identical to the first. Also written text prepared for publishing — both are reproductions of something.
The four basic acts on stored things: create, read, update, delete. A named shorthand rather than a concept.
A plain format holding rows of values separated by commas. Simple enough to outlive every tool that reads it.
Short form of count, and of Connecticut. A cut fragment as it appears in this library, not an English word.
Short form of context — what is woven alongside a thing and gives it meaning. A named abbreviation.
The things that are given. Facts handed over as the ground a reckoning stands on, before anything is concluded from them.
A structured store built so things can be found again reliably. Its whole discipline is retrieval.
To remove repeated copies so each thing is counted once. Deduplication is how an archive stops lying about its size.
To repeat data on purpose so it is faster to read. Denormalising trades one truth for many copies of it.
Short form of dictionary — values reached by name rather than by position. A named structure rather than a concept.
To bring a held thing from elsewhere onto one's own ground. The crossing that makes a copy yours.
A second copy of the same thing. Duplicates inflate a count and are the commonest way an archive lies about its size.
Short form of elements. A cut fragment as it appears in this library, not an English word.
To send outward in a chosen form — render, encode, export, publish. The gateway's verb: how a creation leaves the studio.
Short form of entity — a thing that is, named so it can be counted. A cut fragment as it appears in this library.
A closed set of permitted values, named one by one. Its meaning is as much in what it excludes as in what it lists.
To send a held thing out in a form others can use. Sovereignty over data begins with being able to export it.
An open bounded area, and one named slot within a record. Both are ground marked off so something can be grown or held there.
The name by which a stored thing is known and found. In this library, the filename is the parse.
Structured data set at the head of a document, describing it. Frontmatter is for machines; the body is for readers.
Short form of GEDCOM, the genealogy exchange format. A cut fragment as it appears in this library.
The standard format for exchanging genealogical records. A named format rather than a concept.
Nodes joined by edges, where the joining is the structure. A graph's meaning lives in its connections.
A short fixed value derived from a whole, changing entirely if any part changes. A hash proves sameness without revealing contents.
Base-sixteen notation, used to write colours and bytes compactly. A named encoding rather than a concept.
Ordered guides to where things are. Plural of index — addresses rather than contents.
The taking in of material from outside a system. What is ingested is not yet understood — only held.
What is given to a process from outside it. What a system receives before it can answer.
To bring two things together so they hold, and to become part of a body. Both make one where there were two.
A plain text format for structured data, readable by both machines and people. A named format rather than a concept.
A format holding one record per line, so a file can be read a piece at a time. A named format rather than a concept.
A declared list of what is included. A manifest is a claim about contents, checkable against them.
Established correspondences between two sets. A mapping carries nothing across but the relation itself.
A grid of values addressed by row and column. Its older sense is womb — the ground something forms within.
A compound fragment meaning the highest index. Not an English word; flagged for the mis-cut band.
Data about data. What lets a thing be found, dated, credited and trusted, without being the thing itself.
Contained within another of the same kind. Nesting is depth made of repetition.
A compound fragment meaning a name index. Not an English word; flagged for the mis-cut band.
To bring differing things into one common form. Normalisation makes comparison possible and loses what was particular.
Permitted to hold nothing. Nullability makes absence a valid state rather than an error.
A thing considered as a thing, apart from what it does. Also whatever an action is aimed at.
What a process gives back. The far side of input, and the only part anyone else sees.
To write over what was there, losing it. Overwriting is the one edit that cannot be undone by reading.
To divide a long body into pages that can be asked for one at a time. It makes the unmanageable reachable.
The division of a long body into pages that can be asked for one at a time. It makes the unmanageable reachable.
A named value passed in to shape how something runs. Parameters are where a general thing becomes particular.
That which another descends from or is contained within. The relation is named from above.
The carried content, as against the carrying. The part that matters once the envelope is discarded.
To fill an empty structure with its contents. Populating turns a shape into a thing.
A characteristic belonging to a thing. Also what is owned — the same word for what a thing is and what is held.
A question put to a system in a form it can answer. A query's shape decides what can come back.
Unprocessed, as it arrived. Raw material is the honest ground everything refined must be traceable back to.
To make less, or to bring many down to one. Reduction always discards, and what it discards should be named.
The bringing back of what was stored. Retrieval is the whole reason keeping is done at all.
To bring back what was stored. Retrieval assumes the keeping was done well enough to find it again.
Able to be found by asking. Searchability is a property of how a thing was kept, not of the thing.
One part cut from a longer whole. A segment's boundaries are a decision someone made.
Short form of sequence — things in a fixed order where the order carries meaning. A named abbreviation.
Things in a fixed order where the order carries meaning. Change the order and you change the thing.
To flatten a structured thing into a sequence that can be stored or sent. Serialisation trades shape for portability.
One sharing a parent with another. Siblings are equals by position, which is what distinguishes them from a hierarchy.
A state captured at one moment and held unchanged. A snapshot is true of a time and stops being true immediately.
To put in order by some property. Sorting changes nothing about the things — only where each one stands.
To divide one into parts. A split decides where the seam goes, and the seam is never neutral.
The language for asking structured questions of stored data. A named language rather than a concept.
Something arriving continuously rather than all at once. A stream has no total until it stops.
A run of characters treated as one value. Also a thread — the older sense the data one is named from.
Runs taken from within a longer sequence, keeping their order. A subsequence is a selection that respects the original arrangement.
A flat surface things are set out on, and the ordered arrangement set out there. First a plank cut for inscribing.
The mark for a table header cell. A cut fragment as it appears in this library, not an English word.
A branching structure where every part has one parent and may have many children. It grows outward and never rejoins.
To cut a thing short at a limit. Truncation loses the end and rarely says how much.
The cutting short of a thing at a limit. Truncation loses the end, and honest systems say how much.
To adjust until resonant — an instrument to true pitch, a sound to its intended shape. The bringing of a thing into agreement with itself.
To send a held thing outward into a system that keeps it. The crossing from one's own ground to another's.
Short form of variables — named places whose contents may change. The name stays; what it holds does not.