Result limits and spill files

In DataMind Oda, every step returns a result: a command's output, the matches a search found, the text of a page Oda read. Each result is kept within a size a chat can carry, so a long result arrives shortened, with a short note counting what was left out, and when even the shortened form would not fit, the full result is kept in a file on your computer. This page is the reference for those limits: the exact notes, what each kind of result keeps, and what the stored copy holds.

Why a result gets shortened

A chat carries a limited amount of text. Each tool caps what it returns from a step, so one long result cannot fill the whole chat; the part that does not fit is counted in a short note instead of being silently dropped.
Two kinds of limit. A list of findings, such as search matches or sources, keeps a number of entries; text is kept within a size, measured in bytes.
A limit of zero keeps nothing. When a limit is set to zero, the tool keeps none of that result and counts the whole of it as left out.
Reaching a limit is not a loss. A shortened result reports that a size limit was reached; it never means the source was incomplete, that data was lost, or that the step failed.
Refusals are not omissions. When a step refuses or skips something of its own, such as a file it may not read, it reports that in its own words; it is never counted in an omission note.

The notes that count what was left out

A shortened result carries a short note in place of what was left out:

A list counts entries. When the result is a list of findings, the note reads Omitted 3 items., and the number is how many entries were left out.
Text counts size. When the result is text, the note reads Omitted 12 bytes., and the number is the size of the part left out.
A count without a number. When an exact count is not available, the note carries no number and reads More lines were omitted.
Nothing left out, no note. When a result fits whole, no note appears at all.
The count can join the step's advice. A note may continue past the count with the step's own words on seeing more, for example: Omitted 25 items. Results capped at 100. Narrow the pattern, path, or include to see more. The count is Oda's shared wording; the advice comes from the step, and either half can be missing.
The number is what was actually left out. Cuts first adjust to whole characters, so the count can differ slightly from the limit that caused it.

What each kind of result keeps

The part that stays depends on the kind of result:

Kind of result	What stays
A file search	The first matches it found
A text search	The first matches it found
A web search	The first sources it found
A fetched page	The start of the page; a very large page keeps its start and its end
A command's output	The end of the output; a very large output keeps its start and its end

Cuts fall between characters. When a cut would land inside a character, that character is left out whole rather than broken, and nothing is invented in its place.
A file read is not shortened this way. A long file comes back in parts, and each part says where to continue; Read, write and edit files covers those windows.
The searches are covered in Finding files and text, the notes a command's result carries in Shell commands, and the web steps and their result cards in Web results in chat.

Where the full result is stored

The largest results are kept whole. When even the shortened form would not fit, Oda keeps the full result in a file on your computer; the chat shows a short preview and a note that names the file after Full formatted result stored at:, with the hint Use read with offset/limit, or grep this path for working through it.
The stored copy holds the full formatted result, exactly as it was formatted for the chat; for a fetched page it also keeps the line naming the page and its HTTP status.
Stored copies live outside the chat's working folder. Each chat's stored results sit together in a folder of their own, apart from the folder the chat works in; nothing clears them with the chat, they stay until the start-up sweep ages them out, and a step given a stored path reads or searches that file where it is.
Where the file lives, and when it is cleared up, is covered in Large results and spill files.

Note

The note in the chat is the pointer to the stored copy. Ask Oda to read the file, or to search inside it, when you need the parts that were left out.

Questions, answered

What happens when a limit is set to zero?

That limit keeps none of the result and counts the whole of it as left out. Nothing about the step fails: the note reports that a size limit was reached, the same as any other limit.

Why do I see the first matches but not the rest?

Searches keep the start of their findings: a file search and a text search keep the first matches, and a web search keeps the first sources. What was left out is counted in the note; to see more, narrow the search so a smaller, complete result can come back, as the note's advice suggests.

Why do very large results keep both their start and their end?

For a very large result, both ends carry the most useful information: the start says how it opens and the end says how it finishes. When a fetched page or a command's output grows past its limit, Oda keeps both ends and counts what was left out of the middle.

What does the stored full result hold?

The complete result as it was formatted for the chat; for a fetched page, it also keeps the line naming the page and its HTTP status. The chat itself keeps only a short preview and the note, and Oda reads the stored file back, in parts or by searching inside it, when a task needs the full text.

What does "Results capped at 100" mean?

It is part of the step's own wording in the note: the step keeps 100 entries at most, so everything beyond that was left out and counted in the same note. The advice after it says how to narrow the request so a smaller, complete result can come back.

Why does the advice after the count differ between steps?

Oda's shared note supplies the count; the advice that follows comes from the step itself, in its own words. A search suggests narrowing the pattern, path or filter; another step may say something else, and either half of the note can be missing.

Is the omitted count exact?

Where a number appears, it counts what was actually left out. Cuts are adjusted to whole characters first, so the number can differ slightly from the limit that caused the cut. When no exact count is available, the note carries no number at all and reads More lines were omitted.

Are skipped files counted as omissions?

No. Omission notes count what a size limit left out of a result; they never report refusals or problems. When a step refuses or skips something of its own, such as a file it may not read, it says so in its own words, and that is not counted as an omission.