I hate sub reports and always consider them the last resort in any reporting solution. The negative effect on performance and maintainability is just not worth the easy ride they give the report writer. Nine times out of ten reporting requirements can be met using a little forethought and planning (and a solid understanding of formulas).
That said, there are a few novel ways of using sub reports which will not affect performance and actually prove a boon to the developer.
1. Report Header
Any information, graphics, logos or special fields (Date report was run etc) which will appear in every report can be built into a sub report which is then added to the main report.
The performance hit is minimal, and a small amount is shaved off the development time, plus, it can go a long way to standardising your reports. But the real benefit comes when the business decides to update its logo or corporate color etc. As long as the sub report is set to "Re-import When Opening" (via the sub report's Format Editor), only one report needs to be changed to have impact across the entire report library.
2. Reconciling Conflicting Groups
Often there is a requirement to show the same information summarised by logically conflicting groups. For example: showing the total sales for each week within a month and totals sales per team in a month.
A typical sub report can be used to load the data again then group it by the second value, and this is the typical way to use a sub report. But accessing the database again for data you have is a waste of resources which can be crippling with bigger reports.
The most efficient way to handle this is to load the information you want into one or more arrays and pass them through to the sub report to format and group as you want.
It is possible to display the array in the main report and forgo the need for a sub report at all, but if you are reporting against a lot of data there is a chance the report will finish before the array has been fully displayed.
3. Conditional Data Targets
I come across this issue quite often: a report is needed which always shows the same set of data, plus one of two (or more) other sets of data depending on the user's choice or the results returned from the first set of data.
Because a single report can only have one set of linked tables, multiple sub reports must be used.
For example: a sales report shows revenue for a particular office, if the office has met its target the managers want to see how they compare to the rest of the other offices, but if they fail to meet their target they want to see the sales broken down by each rep to identify any problem areas.
A report based on sales reps and one based on nation office sales require completely different tables. The most efficient way to solve this problem is to create a sub report for each. Whichever is not needed is suppressed and given Record Selection criteria which will return an empty report. The required sub report runs as normal.
A sub report makes the same demands on a database and network resources as a 'proper' report. So a Crystal Report with two to three sub reports can take three to four times longer to run. With this overhead it is usually better to find another solution.
So sub reports can kill a report's performance, but when used with a little imagination they can be a helpful tool in expanding Crystal Reports functionality in a way that cannot be realised by any other method.