JMP was designed and intended for interacting between the program, graphs and the data. Thus, JMP stores everything in memory.
The physical size of the file is not a true indicator of the amount of RAM needed to store the data. For example, suppose you have a data table with only 1 numeric column and 1 million rows. The amount of RAM needed to just hold the data would be calculated as 10 * 1,000,000 = 10,000,000 or 10MB. Character data is calculated by n bytes per character cell, where n is the length of the largest character value in that column.
To determine the size of a data table, the rules of thumb are:
- all numeric values will be stored in 4 bytes (32 bits).
- all character fields will be the length of the longest entry in the column. For example, if most rows are "Yes" and "No" (3 and 2 bytes) but one entry is "Maybe" (5 bytes), all of the rows will have space for the 5 byte entry. Recoding the data could help this situation, such as using a numeric or letter code and making use of the Value Labels table attribute.
- some minor overhead is required for each column and for the data table itself.
Also make note that every data table in use is in memory, and some operations will need memory to operate. For example, having two large data tables already in memory and then trying to merge or concatenate them may result in a new combined table which might exceed available memory. Another possible scenario is attempting to do a large analysis (regression or correlation) on a data table which nearly fills existing memory, leaving no room for the analysis to do the work.
[Previously JMP Note 31493]